Tag: Humanoid Robots & Drones

Hospitality, Leisure and Restaurant Robots and Drones in 2026: The $186 Million Sweetgreen-Wonder Deal That Finally Validated the Category

On December 29, 2025, Los Angeles-based fast-casual salad chain Sweetgreen, Inc. (NYSE:SG) completed the sale of its automated kitchen technology subsidiary Spyce Food, Co. to Wonder Group, Inc. for approximately $186.4 million in combined cash ($100 million) and Series C Preferred Stock ($86.4 million). The transaction transferred ownership of the Infinite Kitchen — the robotic salad-bowl assembly platform that has been the most operationally successful restaurant automation deployment in the United States over the 2023-2025 window — from Sweetgreen to Wonder, which had previously acquired food-delivery operator Grubhub for $650 million in late 2024 and meal-kit pioneer Blue Apron in 2023, and which now operates approximately 80 food-hall locations as it builds what its leadership publicly describes as “a tech-driven food platform owning both robotics and infrastructure.” Sweetgreen had originally acquired Spyce in 2021 for approximately $70 million, including post-acquisition true-up and milestone amounts; the 38 Spyce employees, including cofounders Michael Farid, Kale Rogers, Brady Knight, and Luke Schlueter — all MIT graduates who had built the original Spyce two-unit Boston robot restaurant before the Sweetgreen acquisition — transferred to Wonder as part of the transaction. Sweetgreen retained access to the Infinite Kitchen platform under a long-term supply and services agreement, with plans to continue rolling out automated makelines across approximately half of its 15-to-20 net new restaurant openings in 2026.

The Sweetgreen-Wonder Spyce transaction is the single most consequential commercial validation of restaurant robotics that the industry has produced in its approximately fifteen-year venture-capital investment cycle, parallel in some respects to the recent strategic acquisitions reshaping the broader commercial humanoid robotics landscape. Before December 2025, the restaurant robotics category was best known for its high-profile failures: Zume Pizza, the SoftBank-backed mobile pizza-baking truck operator that burned through more than $445 million in raised capital before pivoting to packaging and ultimately shutting down operations in 2023; Cafe X, the robot-barista kiosk operator that closed all of its San Francisco and Texas locations during the pandemic in 2020; Creator (formerly Momentum Machines), the San Francisco gourmet-burger-robot restaurant that pivoted away from its founding concept; Dishcraft Robotics, the dishwashing-automation specialist that shut down operations in 2022; Pazzi, the Paris-based robot-pizza-restaurant operator that ceased operations in 2022. The category’s commercial trajectory had, until the Sweetgreen-Wonder transaction, looked structurally similar to the collapsed European eVTOL cohort — substantial venture capital deployment, sophisticated engineering, and accumulating partial-success demonstrations that never converted into the operational scaling the original investment thesis required. The Wonder acquisition, at nearly three times the price Sweetgreen had paid for Spyce four years earlier, represents the first genuine commercial validation that a restaurant automation business can produce the unit-economics improvement and operational scaling that makes acquisition by a strategic platform operator economically defensible.

The Infinite Kitchen operational specifics

The Infinite Kitchen makeline is, in mechanical terms, a conveyor-belt-based modular automation platform that dispenses pre-measured ingredients through controlled hoppers into individual customer salad bowls as the bowls travel along a continuous belt, with the dispensing sequence driven by the digital point-of-sale order data and with the final assembly stage (final mixing, dressing application, garnish placement) performed by human team members at the end of the line. The platform operates at throughput of approximately 400 to 500 bowls per hour, against the approximately 150-200 bowls per hour that a traditional Sweetgreen makeline operates at, while requiring approximately half the front-line labor headcount of a comparable conventional store. Sweetgreen’s publicly-disclosed unit economics improvement at Infinite Kitchen locations runs at approximately 700 basis points (7 percentage points) of labor savings against comparable-vintage conventional locations and approximately 100 basis points of cost-of-goods-sold improvement, driven primarily by reduced portion-control variability. The first Infinite Kitchen location opened in Naperville, Illinois on May 10, 2023. The 20-plus-store installed base as of late 2025 includes deployments across California, the Midwest, the Northeast, and the company’s first drive-thru-plus-Infinite-Kitchen format “Sweetlane” location in Costa Mesa, California.

The Spyce technology trajectory — from MIT undergraduate project to 2018 Boston restaurant launch to 2021 Sweetgreen acquisition to 2023 Infinite Kitchen commercial launch to 2025 Wonder acquisition at nearly 3x the original purchase price — is the cleanest available case study of how a restaurant automation business actually achieves commercial validation. The four cofounders’ academic robotics genealogy at MIT anchored the technology development in fundamentals research rather than the pure venture-investment-and-marketing model that characterized many of the failed restaurant robotics platforms of the late 2010s. Wonder’s broader strategic platform — combining the Spyce in-restaurant kitchen automation with the Grubhub delivery infrastructure and Blue Apron meal-kit fulfillment under a unified operational architecture — represents a thesis about the integrated economics of food production, distribution, and last-mile logistics that no other operator in the restaurant industry has assembled at comparable scale.

Miso Robotics Flippy: White Castle, Jack in the Box, and the CaliExpress all-robot quick-service launch

The longest-deployed-into-commercial-operation restaurant robotics platform in the United States is Miso Robotics’ Flippy, the autonomous fry-station and burger-grill robot that the Pasadena-based company has been refining through multiple product generations since 2017. Flippy has been operationally deployed at CaliBurger restaurants since 2017, at White Castle locations beginning with the Merrillville, Indiana store in 2020 and expanding to additional U.S. locations over the subsequent four years, and at Jack in the Box locations beginning with the company’s 2022 announced partnership. The Flippy 2 platform automates the fry station — taking frozen french fries from the freezer, placing them in the fryer, monitoring cook time, removing them at the correct doneness, salting them, and placing them in the hot-hold position — at throughput equivalent to a human fry cook but with lower variability in cook time and salting consistency.

In January 2024, Miso Robotics launched CaliExpress by Flippy in Pasadena, California — the first commercially-operating fully-autonomous fast-food restaurant in the United States, in which Flippy operates the fry stations, additional robotic systems operate the burger grills, and Cecilia.AI‘s robotic bartender mixes the drinks. The CaliExpress format was positioned by Miso CEO Rich Hull as the operational demonstration of what an all-robot fast-food restaurant could actually look like at unit-economics scale, rather than as a primary growth vehicle for the company. The commercial customer pipeline — White Castle, Jack in the Box, Inspire Brands’ Buffalo Wild Wings — remains the core revenue model. Miso has, over the 2018-2025 window, raised approximately $108 million in disclosed venture capital across multiple rounds, with deployed-Flippy unit counts in the low hundreds across the company’s commercial customer base.

Chipotle Autocado, Augmented Makeline, and the legacy-chain robotics integration story

The largest single restaurant chain executing a public robotics deployment program in 2026 is Chipotle Mexican Grill (NYSE:CMG), under former CEO Brian Niccol (who departed for Starbucks in August 2024) and his successor Scott Boatwright. Chipotle’s robotics initiatives include the Autocado — an avocado-processing robot developed in partnership with Pasadena-based Vebu Labs that automates the cutting, pitting, and scooping of avocados for guacamole preparation, reducing the time required to prep a batch of guacamole from approximately 50 minutes to approximately 25 minutes — and the Augmented Makeline, an automated bowl-and-salad assembly platform developed in partnership with Hyphen (the South San Francisco automation company that builds back-of-house automation platforms for ghost kitchens and traditional restaurants). The Autocado was first deployed in test at the Chipotle innovation center in Irvine, California in 2023 and has been progressively rolled out to additional locations over 2024-2025. The Augmented Makeline addresses the digital-channel order assembly bottleneck that has, in Chipotle’s documented operational reporting, consumed disproportionate front-of-house labor as the company’s mobile-app-and-digital-channel ordering has grown to more than 35 percent of total sales.

The Chipotle robotics deployment strategy reflects the legacy-chain operational logic: the platform must integrate into existing restaurant footprints, must improve specific high-labor-cost operations rather than replacing entire kitchen operations wholesale, and must produce measurable per-store ROI within capital-payback timeframes that the company’s financial planning process will support. The strategy is structurally different from the Sweetgreen-Spyce or CaliExpress all-robot integrated approach, in which entire restaurant formats are designed around the automation platform from the ground up. Both approaches have produced operationally successful deployments. The legacy-chain integration approach, by deployed-unit count, will produce the larger total robotic footprint over the next five years simply because Chipotle’s 3,500-plus North American store base substantially exceeds the combined footprint of every dedicated all-robot restaurant operator currently in the market, in operational parallel to the logistics warehouse robotics deployment pattern where legacy operators retrofitting existing facilities have produced the larger total robotic installed base than greenfield all-robotic warehouses.

Kernel by Steve Ells: the Chipotle founder’s NYC robotic restaurant

In April 2024, Steve Ells — the founder of Chipotle Mexican Grill, who departed the CEO role in 2017 — opened Kernel in New York City’s Greenwich Village neighborhood, with subsequent locations opening across Manhattan through 2024 and 2025. Kernel operates as an extensively-automated fast-casual restaurant built around a Mitsubishi articulated industrial robot arm that handles the food assembly and tray-loading operations for the company’s plant-based menu, with a substantially smaller human staff than a conventional Kernel-sized restaurant would require. The Kernel concept is, in operational positioning terms, the highest-profile post-Chipotle restaurant robotics launch from one of the most operationally successful fast-casual restaurant operators of the past three decades. Whether Kernel scales beyond its current Manhattan footprint into a national chain — as Chipotle did from its first Denver location in 1993 — will be one of the more consequential commercial signals for restaurant robotics over the 2026-2030 window.

The in-restaurant delivery robot category: Bear Robotics Servi, Pudu BellaBot, Keenon

The most visible — though, by operational impact, less consequential — restaurant robotics category in 2026 is the in-restaurant delivery robot, the small wheeled platform that carries plates from the kitchen to dining room tables. The category-leading platform is Bear Robotics’ Servi, the in-restaurant delivery robot developed by the Redwood City, California-based company founded in 2017 by John Ha (a former Google engineer and Korean restaurant owner). Bear Robotics raised a $60 million Series C funding round in 2024 led by LG Electronics, with Naver and SoftBank as additional strategic investors, valuing the company at approximately $400 million. Servi has been deployed across major U.S. casual-dining chains including Denny’s, Chili’s, and Cracker Barrel, with deployed-unit counts in the low five figures. The Servi platform is fundamentally a wheeled tray-carrying mobile robot with obstacle avoidance, optimized to navigate the heterogeneous obstacle environment of a restaurant dining room with seated customers, moving servers, and unpredictable foot traffic.

The Chinese competing platforms — Pudu Robotics’ BellaBot (the cat-faced delivery robot with the animated facial display that has become the most photographed restaurant robot in the world), and Keenon Robotics’ DINERBOT — operate in the same category at lower price points, with the BellaBot in particular having achieved global deployment across thousands of restaurants in dozens of countries. The category leader by deployed-unit count globally is BellaBot. The category leader by revenue-per-unit in the U.S. is Bear Robotics’ Servi.

The bartending robot category: Richtech Robotics ADAM, Cecilia.AI, and the Royal Caribbean Bionic Bar

The bartending robotics category has, since approximately 2019, been the most operationally theatrical subcategory of restaurant robotics — the robot bartender platforms designed primarily for the spectacle of drink preparation in front of customers rather than for labor-cost reduction. Richtech Robotics (NASDAQ:RR), the Las Vegas-based robotics company that completed its initial public offering in November 2023, operates the ADAM dual-arm bartending and barista robot platform. ADAM units have been deployed across hotels, conference centers, sports venues, and casino properties, with the Richtech operational footprint expanding into the broader food-service automation category. Cecilia.AI, the robotic bartender platform that operates at CaliExpress by Flippy, occupies the same operational niche at smaller scale.

The most operationally-scaled bartending robot deployment in the world is the Bionic Bar aboard the Royal Caribbean International cruise fleet. The Bionic Bar — a fully-automated bar staffed by two ABB IRB 2600 articulated industrial robotic arms that mix and serve cocktails to passengers — operates across at least five Royal Caribbean ships: Quantum of the Seas (introduced 2014), Anthem of the Seas, Ovation of the Seas, Odyssey of the Seas, and Wonder of the Seas. The Bionic Bar concept is structurally a piece of cruise-ship entertainment theater rather than a labor-cost-reduction deployment — the robots mix drinks at a pace slower than experienced human bartenders, but the visual spectacle of watching ABB industrial arms perform choreographed cocktail mixing is the experiential product. The Bionic Bar has, by every available indication, been operationally successful as entertainment theater, with sustained passenger engagement across the decade since its 2014 introduction, in operational contrast to the bolted-down ABB articulated arms running automotive and construction-site welding operations where the same hardware platform exists strictly to perform repetitive industrial labor outside of public visibility.

The hotel robotics category: Henn-na Hotel’s reset, Savioke Relay, and the operational limits of front-of-house service automation

The hotel robotics category in 2026 is fundamentally smaller and less operationally successful than the restaurant or cruise ship categories. The most famously aggressive hotel automation deployment — the Henn-na Hotel in Sasebo, Japan, which opened in 2015 with an explicitly “all-robot staff” marketing positioning including humanoid robot reception, robotic luggage handlers, and in-room virtual assistants — publicly fired approximately half of its 243 robots in 2019 after persistent reports of operational failures, including in-room voice assistants triggered by overnight snoring, baggage robots that failed in outdoor temperatures, and reception robots that could not handle non-routine guest questions. The Henn-na reset became one of the most-cited operational case studies of the limits of front-of-house service automation in hospitality contexts. The hotel chain continues to operate multiple locations in Japan with a substantially reduced robotic footprint.

The most operationally successful hotel robotics deployment in the United States was the Savioke Relay in-hotel delivery robot, developed by the San Jose-based Savioke (later acquired by Relay Robotics in 2021) and deployed across Aloft Hotels properties and additional hotel chains beginning in 2014. The Relay platform delivers small items — toothbrushes, towels, snacks, room service items — from the front desk to guest rooms, using elevator integration and floor-mapping autonomy. The deployment has been operationally stable but commercially modest, with deployed-unit counts in the low hundreds globally. SoftBank Robotics’ Pepper humanoid was deployed at concierge positions across multiple hotel chains in the mid-2010s; most of those deployments have been progressively wound down as the Pepper platform’s operational limitations in unstructured guest-interaction scenarios became evident, paralleling the broader pattern observed in adjacent eldercare-and-hospitality humanoid deployments where Pepper found similar adoption limits.

The Disney Imagineering BDX droids, Stuntronics, and the theme park animatronics genealogy

The most operationally sophisticated entertainment robotics platforms in the world operate inside The Walt Disney Company’s theme park operations. Disney Imagineering‘s BDX droids — the rolling, expressive, autonomous service-droid platforms that interact with guests in Star Wars: Galaxy’s Edge at Disneyland and Disney’s Hollywood Studios — have operated as guest-facing entertainment characters since 2023, building on Disney Imagineering’s broader autonomous-character development program that traces back to Project Kiwi (the public name for the company’s free-roaming-character development initiative). Disney Imagineering’s Stuntronics platform, publicly demonstrated in 2018, is an autonomous trapeze-stunt humanoid robot capable of executing aerial acrobatic sequences without human safety wires. The Stuntronics platform is used inside Disney’s Marvel-themed attractions for autonomous superhero-character flight sequences. The Disney animatronics genealogy traces back to the Tiki Room attraction at Disneyland in 1963 — sixty-three years of continuous entertainment robotics deployment that no other organization in the world matches, predating the modern quadrupedal robot category by approximately a half century.

The competing theme-park robotics deployment is at Universal Studios (Comcast) and at the Disneyland Paris, Tokyo Disney, and Shanghai Disney facilities. Stuntronics-class autonomous stunt robotics has been replicated to limited extent at Universal’s Marvel-and-Wizarding-World attractions, though Disney Imagineering’s lead in the autonomous-character category remains substantial.

The drone-show category at resorts, casinos, and theme parks

The drone-show category — coordinated swarms of 100 to 1,000-plus drones executing choreographed aerial light displays — has become the entertainment-robotics technology that has scaled most rapidly across hospitality venues in 2024-2026. The category-leading operators include Verge Aero (Philadelphia-based, the drone-show specialist that operates the Disney drone shows and major Las Vegas resort productions), Sky Elements Drone Shows (Fort Worth-based, holder of multiple Guinness World Records for largest drone-show formations), and Pixis Drones (Las Vegas-based, focused on casino and convention deployments) operates within the same Las Vegas hospitality ecosystem where casino surveillance and security technology has historically been the most operationally sophisticated commercial deployment of detection and identification systems anywhere in the world. Major hospitality drone-show deployments include the Sphere in Las Vegas, Disney Springs, multiple Las Vegas casino properties including Caesars Palace and Wynn Las Vegas, and large-scale resort installations at Atlantis Bahamas, Sandals Resorts properties, and major casino-resort operations across Macau and Singapore, with the underlying drone swarm coordination infrastructure now overlapping operationally with the emergency-response drone technology used by fire and disaster agencies. The drone-show category is, by operational definition, a coordinated swarm autonomy application — each individual drone executes a pre-programmed flight path with onboard GPS-RTK positioning and LED color sequencing, with a central ground-station orchestrating the swarm-level choreography. The category has, over the 2022-2026 window, substantially displaced traditional pyrotechnic fireworks displays at major hospitality and entertainment venues, driven by lower environmental-impact regulations, reduced fire-risk concerns (particularly in drought-prone Western U.S. resort and theme-park locations where wildfire-mitigation policy now actively discourages pyrotechnics), and the visual capability for far more elaborate choreographed sequences than conventional fireworks can produce.

What 2026 looks like across hospitality, leisure, and restaurant robotics

In 2026, the hospitality and restaurant robotics category is structurally dominated by a small number of operationally validated platforms in each subcategory. Restaurant kitchen automation is dominated by the Wonder-owned Spyce Infinite Kitchen (20+ Sweetgreen locations, expanding to half of Sweetgreen’s 2026 new openings), Miso Robotics’ Flippy (deployed at White Castle, Jack in the Box, CaliBurger, and the all-robot CaliExpress demonstration restaurant), Chipotle’s Autocado and Augmented Makeline (in progressive rollout across the chain’s 3,500-plus locations), and Steve Ells’ Kernel concept (expanding from the original NYC Greenwich Village location across Manhattan). In-restaurant delivery is dominated by Bear Robotics’ Servi (Denny’s, Chili’s, Cracker Barrel deployments backed by LG Electronics, Naver, and SoftBank) and the Chinese competing platforms led by Pudu BellaBot and Keenon DINERBOT. Bartending robotics is dominated by Richtech Robotics’ ADAM and the Royal Caribbean Bionic Bar (ABB IRB 2600 arms on five-plus ships). Hotel robotics has retrenched substantially from its 2015-2019 expansion peak, with the Henn-na Hotel reset and the wind-down of SoftBank Pepper hotel deployments representing the operational reality that front-of-house guest-interaction automation has not yet produced the reliability that hotel operators require. Theme park robotics is dominated by Disney Imagineering’s BDX droids, Stuntronics platforms, and the broader sixty-three-year animatronics genealogy. Drone-show entertainment is dominated by Verge Aero, Sky Elements, and Pixis at the largest casino, theme park, and resort venues globally.

The structural story across hospitality robotics in 2026 is the bifurcation between operationally successful back-of-house labor automation (Sweetgreen Infinite Kitchen, Chipotle Autocado, Miso Flippy at QSR chains) and operationally successful front-of-house entertainment automation (Disney BDX, Royal Caribbean Bionic Bar, Verge Aero drone shows), with the front-of-house guest-service automation category in the middle — the Pepper concierges, the Henn-na Hotel robot receptionists, the early Bear Robotics initial deployments — having largely failed to produce the operational reliability and customer experience that hotel and restaurant operators required. The two ends of the spectrum work for different reasons. Back-of-house automation works because the operational task is structured, the labor-cost reduction is measurable, and the customer never directly interacts with the robot. Entertainment automation works because the operational task is choreographed, the customer experience is the product (rather than incidental to it), and the unpredictability that destabilizes front-of-house service automation is absent. The middle category — guest-facing service automation in unstructured interaction contexts — remains the operationally hardest category in hospitality robotics, and remains underdeveloped against the original 2015-2019 industry expectations, paralleling the similar adoption challenges for patient-facing service robotics in healthcare settings where unstructured human-interaction contexts have similarly resisted full automation.

The Sweetgreen-Wonder Spyce transaction at $186.4 million in December 2025 establishes the operational and financial validation that the broader restaurant robotics category had spent fifteen years trying to produce. Wonder’s strategic thesis — building a tech-driven food platform combining Grubhub delivery infrastructure, Blue Apron meal-kit fulfillment, and Spyce robotic kitchen automation into a unified operational platform — represents the first major consolidation of the restaurant robotics category by a strategic-platform acquirer rather than the founder-led venture-capital-funded growth model that characterized the previous decade. Whether Wonder ultimately scales the integrated platform into the meaningful commercial-revenue franchise the acquisition thesis implies is the question that will define the 2026-2030 commercial trajectory of restaurant robotics. The signals from the Sweetgreen Infinite Kitchen deployment — 700 basis points of labor savings, 100 basis points of COGS improvement, 400-500 bowls per hour throughput against 150-200 in conventional makelines, half the staffing requirement — are the operational data points that the broader industry will be referring back to as it makes the capital deployment decisions that determine which restaurant robotics platforms scale and which follow Zume Pizza, Cafe X, Creator, Dishcraft, and Pazzi into the operational casualty list.

The robots that successfully populate the hospitality and restaurant industry in 2026 are not the robots the 2015-era restaurant-of-the-future marketing campaigns predicted. They are not humanoid platforms standing behind counters taking orders. They are not robot bartenders mixing drinks faster than humans. They are not robot concierges greeting hotel guests. They are conveyor-belt assembly platforms dropping pre-measured ingredients into bowls at controlled portions, articulated arms cycling french fries through the fryer at consistent doneness, autonomous wheeled platforms carrying plates between kitchens and dining tables, ABB industrial arms performing choreographed cocktail-mixing for cruise passengers, and Disney Imagineering animatronic characters that have evolved across six decades of continuous deployment. The operational logic is the same logic that has driven robotic deployment in factories, warehouses, hospitals, construction sites, and the broader industrial economy: pick a single repetitive task that has high labor cost, low task complexity, and structured operational context; automate that task with hardware purpose-built for the application; let humans handle everything else. The Infinite Kitchen at the Sweetgreen Naperville store dispenses ingredients into bowls. It does not greet customers, mix dressings, garnish plates, clear tables, run a register, or do any of the other dozen things a conventional Sweetgreen location requires. The five-employee staffing of the Infinite Kitchen format handles all of those tasks, working alongside the robot that handles the assembly bottleneck.

That operational architecture — robot doing the bottleneck task, humans doing everything else — is the architecture that has, in 2026, finally produced the commercial validation that the restaurant robotics category had been chasing since the 2014-2015 venture capital wave first crested. Wonder Group’s $186.4 million investment in the Spyce business is, in operational terms, a bet that the architecture scales. Whether it scales beyond Sweetgreen into the broader restaurant industry — and whether the parallel deployments at Chipotle, Miso’s customer base, and Kernel ultimately produce the multi-hundred-thousand-deployed-unit footprint that the venture capital case originally promised — will be determined by the operational data generated across the next half decade inside the same restaurants, hotels, cruise ships, and theme parks that have, since the early 2010s, been the proving ground for hospitality robotics at every prior moment the category attempted to scale.
Aviation and Airports Robots and Drones in 2026: Inside the Most Heavily Regulated, Slowest-Moving, and Most Operationally Bifurcated Robotics Domain

In November 2025, the Federal Aviation Administration granted Joby Aviation (NYSE:JOBY) — the Santa Cruz, California-based electric vertical takeoff and landing (eVTOL) aircraft developer founded in 2009 by JoeBen Bevirt — Type Inspection Authorization for the company’s five-seat air taxi aircraft, marking the first eVTOL manufacturer in the United States to progress to Stage 4 of the FAA’s five-stage type certification process. The announcement came after Joby completed more than 850 test flights in 2025, surpassing 50,000 total flight miles across operations in the United States, the United Arab Emirates, and Japan — including 41 flights at the World Expo 2025 in Osaka and 21 flights during an environmental and operational testing campaign in the UAE. The Joby aircraft is a 5-seat (1 pilot, 4 passengers) all-electric tilt-rotor design with six electric propulsion units, capable of approximately 200 mph in cruise flight with a target range of 100 miles per charge, intended for short-distance urban and suburban routes. Joby’s strategic partnership with Delta Air Lines, announced in October 2022, positions the company to launch initial commercial operations across New York and Los Angeles markets in late 2026 or early 2027, pending the issuance of the full Type Certificate and Part 135 Air Carrier Certificate. Toyota Motor Corporation has invested approximately $894 million in Joby since 2020, with an additional $250 million commitment announced in 2025, bringing Toyota’s total investment to nearly $1 billion — the single largest strategic-partner investment in any U.S. eVTOL developer.

On March 27, 2026, the FAA published its final airworthiness certification standards for powered-lift aircraft — the regulatory category that covers eVTOL air taxis — in the Federal Register, capping a multi-year rulemaking process that the eVTOL industry had publicly identified as the single largest regulatory barrier to commercial operation. The new framework establishes performance benchmarks, pilot qualification requirements, and maintenance protocols that manufacturers must meet before carrying fare-paying passengers. Joby publicly described the FAA action as a “defining moment for American aviation” and committed to obtaining its full Type Certificate within 12 months of the rule publication. Archer Aviation (NYSE:ACHR), the San Jose-based Joby competitor that has logged more than 400 test flights of its five-seat Midnight aircraft, issued parallel commitments. The path from Type Certificate to actual commercial service, by analyst consensus expressed in late 2025 and early 2026, is now projected at late 2027 or later for the first commercial passenger operations, rather than the earlier 2025-2026 timelines the industry had publicly committed to in 2022 and 2023.

The Joby-Archer survival story is the operational counter-narrative to the eVTOL industry’s most consequential 2024 development: the collapse of the European competitive cohort. The German eVTOL industry, which by 2023 had been considered roughly competitive with the U.S. cohort and ahead of the broader Chinese cohort in terms of certification progress, effectively imploded over the second half of 2024. Lilium GmbH, the Munich-based developer of a ducted-fan electric jet designed for regional air mobility — capable of carrying six passengers across a 150-mile range — filed for insolvency proceedings in October 2024 after burning through approximately $1.5 billion in raised private capital across its lifecycle. The company secured a December 24, 2024 rescue deal with a consortium called Mobile Uplift Corporation, then watched the rescue financing fall through in February 2025, resulting in a second insolvency filing. Volocopter GmbH, the Bruchsal-based developer of the VoloCity two-seat urban air taxi backed by Mercedes-Benz, Geely, and a consortium of European investors, filed for insolvency proceedings at the Karlsruhe Local Court on December 26, 2024 after failing to secure the bridge financing needed to carry the company through to EASA Type Certification. The combined Lilium-Volocopter collapse eliminated the European eVTOL cohort’s two most operationally advanced manufacturers, raising structural questions about whether the American AAM industry can survive without the broader transatlantic competitive ecosystem that had been assumed throughout the 2018-2023 venture-capital-investment wave, in marked contrast to the substantially better-capitalized commercial humanoid robotics industry that has continued to attract major investment rounds across the same time window.

The Chinese eVTOL story: EHang’s October 2023 type certification milestone

The first eVTOL aircraft to receive type certification from any civil aviation authority in the world was the EHang EH216-S — a two-seat autonomous (no pilot on board) urban air taxi developed by the Guangzhou-based EHang Holdings (NASDAQ:EH). The Civil Aviation Administration of China (CAAC) granted EHang’s EH216-S type certification in October 2023, ahead of any equivalent U.S. or European regulatory approval. The EH216-S is an 8-rotor coaxial multicopter with an empty weight of approximately 360 kilograms, designed for short urban hops of up to 30 kilometers at a top speed of 130 kilometers per hour. EHang has subsequently received Production Certificate approval and has flown commercial operational demonstrations in multiple Chinese cities including Hefei, Shenzhen, and Guangzhou. The Chinese eVTOL category, including parallel competitors at AutoFlight, Aerofugia (Zhejiang Geely-backed), and TCab Tech, represents the most operationally advanced regulatory framework for autonomous urban air mobility anywhere in the world as of early 2026 — though Western aviation analysts have raised structural concerns about whether Chinese certification standards meet the safety thresholds applied by the FAA and EASA.

The autonomous fixed-wing aviation category: Reliable Robotics, Merlin Labs, and Joby’s Xwing acquisition

Distinct from the eVTOL category — but operating within the same broader autonomy-in-aviation domain — is the autonomous fixed-wing aircraft category, dominated by a small number of U.S. companies developing software-and-sensor packages that enable existing certified cargo aircraft to operate without on-board pilots. Reliable Robotics, the Mountain View-based company founded in 2017 by former SpaceX engineers Robert Rose and Juerg Frefel, has developed an autonomous flight system installed on the Cessna 208 Caravan cargo turboprop, completing autonomous taxi, takeoff, cruise, and landing flights at the Hollister Municipal Airport in California in November 2023 — the first publicly-documented uncrewed flight of a certified cargo aircraft in U.S. civilian airspace. The company subsequently entered into agreements with FedEx for cargo operations integration and has been pursuing FAA Supplemental Type Certificate approval for the Caravan autonomy retrofit. Merlin Labs, the Boston-based competing autonomous flight company, has developed similar autonomy systems for the Caravan and the Beechcraft King Air, with operational testing at Hawaii’s Mokulele Airlines and other regional cargo carriers. Xwing, the San Carlos-based competing autonomous Cessna Caravan developer, was acquired by Joby Aviation in September 2024 for approximately $200 million in stock, with the Xwing technology and team being absorbed into Joby’s broader autonomy roadmap.

The European autonomous flight equivalent is Airbus’s Autonomous Taxi, Take-Off, Landing (ATTOL) project, which completed approximately 500 autonomous flight cycles between 2019 and 2020 on an A350-1000 test aircraft, demonstrating computer-vision-based autonomous taxi, takeoff, and landing capability without GPS or instrument landing system inputs. Airbus has subsequently absorbed the ATTOL learnings into the broader Wayfinder autonomy program. The autonomous fixed-wing aviation category, in 2026 operational terms, is smaller than the eVTOL category by deployed-unit count but is, by available evidence, closer to actual commercial revenue — Reliable Robotics and Merlin have demonstrated the operational capability that would, with appropriate FAA approvals, enable autonomous cargo operations on existing aircraft types within a regulatory framework that already exists rather than the new powered-lift category that eVTOL operators required.

The counter-UAS market: Dedrone-Axon, DroneShield, Anduril, and the Gatwick legacy

The single largest operationally significant aviation robotics market that the broader public-attention apparatus does not adequately cover is the counter-UAS (unmanned aircraft system) market — the dedicated radar, radio-frequency detection, optical-recognition, and active-countermeasure systems deployed by airports, military bases, and critical infrastructure operators to detect and respond to unauthorized drone activity. The category’s foundational moment was the December 2018 Gatwick Airport drone incident, in which reports of drone activity over the runway forced the closure of the United Kingdom’s second-busiest airport for approximately 36 hours, with an estimated 1,000 flights cancelled and 140,000 passengers affected. The Gatwick incident — for which no perpetrators were ever identified — established that a small number of consumer-grade drones could, with minimal coordination, cause economic damage approaching $50 million to a major hub airport in a single incident. The structural response over the subsequent six years has been the rapid expansion of a dedicated counter-UAS industry.

The category leader in commercial counter-UAS is Dedrone, the Kassel, Germany-based and Reston, Virginia-headquartered company founded in 2014 that builds RF-detection-based counter-UAS systems deployed across airports, stadiums, prisons, military bases, and high-value commercial facilities globally. In October 2024, Axon Enterprise (NASDAQ:AXON) — the publicly-traded developer of police body cameras and the TASER conducted-energy weapon — announced the acquisition of Dedrone for approximately $586 million, integrating Dedrone’s counter-UAS capability into Axon’s broader public-safety platform. DroneShield Limited (ASX:DRO), the Australian counter-UAS specialist, has experienced substantial revenue growth across 2023-2025 driven by U.S. and Australian Department of Defense procurement, with the company’s market capitalization reaching multi-billion-dollar levels by mid-2025. Anduril Industries, the Costa Mesa-based defense technology company founded by Palmer Luckey (the Oculus VR founder), operates the Lattice open architecture command-and-control platform with extensive counter-UAS deployment at U.S. military bases and select critical infrastructure sites, alongside the Anvil interceptor drone designed specifically for counter-UAS kinetic engagement. Echodyne, the Washington-state-based metamaterials radar specialist, builds the EchoGuard and EchoShield ground-based radars used in counter-UAS deployments. Fortem Technologies operates the competing radar-and-AI counter-UAS platform.

The structural challenge of counter-UAS in U.S. civilian airports is the legal authority limitation built into U.S. federal law. Under the 2018 Preventing Emerging Threats Act, only four federal departments — DHS, DOJ, DOE, and DOD — have explicit statutory authority to detect, identify, and disable unauthorized drones in U.S. airspace. State, local, and private-sector counter-UAS operations are, under the existing legal architecture, substantially constrained, even when the detected drone activity directly threatens commercial aviation operations at the operator’s own facility, in operational contrast to the broader state and local law enforcement drone procurement authorities that have expanded substantially since 2018. The late-2024 wave of drone sightings over New Jersey that drew national attention through December 2024 and January 2025 illustrated the legal-architecture limitations: even with extensive federal interagency coordination, the inability of state and local authorities to take direct action against the detected drones was a documented gap that subsequent legislative and regulatory proposals are still working to address as of 2026.

The airport ground robotics category: Brain Corp, Avidbots, and the cleaning fleet expansion

The most operationally scaled robotics category inside actual airport facilities is commercial cleaning robotics — the autonomous floor scrubbers, sweepers, and vacuum systems deployed across airport terminal buildings to maintain the floors, restrooms, and concourses that millions of passengers traverse daily. The category-leading software platform is BrainOS, developed by Brain Corporation in San Diego, which provides the autonomy software stack integrated into floor-care equipment manufactured by Tennant Company (NYSE:TNC), Nilfisk, SoftBank Robotics’ Whiz scrubber, and additional commercial cleaning OEMs. BrainOS-powered cleaning robots are operationally deployed across more than 100 U.S. airports as of 2024 disclosure, including hub airports such as Hartsfield-Jackson Atlanta International, Los Angeles International, Chicago O’Hare, Dallas-Fort Worth, and John F. Kennedy International. The Avidbots Neo autonomous floor scrubber, built by the Kitchener, Ontario-based Avidbots Corporation, operates as a competing platform with significant airport deployment including Cincinnati/Northern Kentucky International and Pittsburgh International. The cleaning robotics category is, in airport operational terms, the most quietly successful single example of back-of-house robotic deployment in the industry — invisible to passengers, transparent to airline operations, and quietly addressing the persistent labor-cost-and-availability problems that airport ground services contractors have struggled with for decades.

The baggage handling automation category and ground services

The baggage handling automation subcategory is dominated by Vanderlande Industries (a subsidiary of Toyota Industries), the Dutch company headquartered in Veghel that builds the conveyor systems, automated sorting carousels, and increasingly the robotic baggage-handling cells deployed in major hub airports globally. Vanderlande’s installed base includes baggage handling at Amsterdam Schiphol, Hong Kong International, Frankfurt, Heathrow Terminal 5, JFK Terminal 4, and dozens of additional major facilities. Daifuku Co., Ltd. (TYO:6383), the Osaka-based competing materials-handling automation specialist, operates parallel baggage automation deployments. Beumer Group and BEUMER Crisplant operate the CrisBag baggage handling automation system deployed across major European airports. The category has been operationally mature for decades but has increasingly incorporated robotic articulated-arm components for the specific tasks of removing baggage from conveyors into containers, with major airport modernization programs at Frankfurt and Amsterdam Schiphol explicitly including robotic baggage-handling cells in the 2023-2026 capital deployment plans.

The autonomous ground services equipment category includes electric autonomous pushback tractors (Mototok, Goldhofer, TLD), autonomous belt loaders for baggage cart loading, and the broader category of autonomous ramp equipment that airports and ground-handling contractors are increasingly investing in as the labor shortage in ground services has paralleled the broader logistics and warehouse worker shortages that have driven adjacent robotics adoption. The category remains, in 2026 operational terms, less mature than airport cleaning robotics or baggage handling automation, primarily because the integration challenges of operating autonomous equipment on active ramps with conventional ground-handling crews remain substantial, in operational contrast to the more controlled environments where autonomous heavy equipment has scaled in mining and resource extraction.

The aircraft inspection drone category

The aircraft inspection drone category — autonomous drones that perform pre-flight visual inspections of commercial aircraft, looking for hail damage, lightning strikes, paint defects, fuselage damage, and other surface anomalies — is dominated by two operationally focused specialists. Donecle, the Toulouse-based French company, builds a fully-autonomous drone that performs laser-scanning and high-resolution photographic inspection of commercial aircraft in approximately 15-20 minutes (versus 4-8 hours for traditional manual inspection on a cherrypicker), with operational deployments at Air France Industries KLM Engineering & Maintenance, Lufthansa Technik, Delta TechOps, Iberia Maintenance, and additional MRO (maintenance, repair, overhaul) facilities globally. Mainblades, the Delft, Netherlands-based competing aircraft inspection drone specialist, operates a similar platform with parallel MRO customer deployment. The aircraft inspection drone category, in operational terms, has demonstrated time-and-cost reductions that align well with the broader MRO sector’s labor and capital constraints, but the category remains small in deployed-unit count against the broader airport-and-aviation robotics market.

The bird control and wildlife management drone category

The bird strike threat to commercial aviation — most famously illustrated by the 2009 US Airways Flight 1549 “Miracle on the Hudson” incident, when a Canada goose collision with both engines forced Captain Chesley “Sully” Sullenberger to ditch the Airbus A320 in the Hudson River — drives an ongoing operational requirement for active wildlife management at every major commercial airport. The Robird, a Dutch-developed bird-shaped drone modeled on a peregrine falcon, manufactured by Clear Flight Solutions in Enschede, Netherlands, operates at airports including Edmonton International and Southampton Airport to deter bird populations through visual mimicry of natural predators. USDA Wildlife Services, which provides bird hazard management at many U.S. airports under interagency agreements, has expanded use of conventional drones (DJI Matrice and Skydio platforms) for wildlife surveillance and harassment alongside traditional methods, with experimental deployment of quadrupedal ground robots from Boston Dynamics and Ghost Robotics for airfield perimeter monitoring at several major U.S. airports. The bird control category is operationally niche but addresses a documented commercial aviation safety risk that the FAA, ICAO, and airport operators have classified as a continuing priority, drawing on the broader wildlife management robotics ecosystem that has expanded across conservation and land-management applications over the past decade.

The eVTOL infrastructure question: vertiports and the integration challenge

The structural question that will determine whether the surviving eVTOL operators — Joby, Archer, Beta Technologies, EHang, Wisk Aero, Eve Air Mobility — actually scale into meaningful commercial service is not aircraft certification but vertiport infrastructure. A commercial eVTOL operation requires dedicated takeoff-and-landing facilities, electric charging infrastructure at scale, passenger ground access, integration with existing helicopter air traffic control protocols, and — for the urban use case that the entire industry’s investment thesis depends on — political acceptance from the residential populations directly underneath the proposed flight paths. Skyports Infrastructure (London-based), Atlantic Aviation, Signature Aviation (now part of Global Infrastructure Partners), and the broader FBO (fixed-base operator) industry are all pursuing vertiport development. Lilium Pad, Volocopter VoloPort, and similar branded vertiport concepts proliferated in the 2021-2023 industry communications environment but have, in 2026 operational reality, produced relatively few operationally completed facilities. The dependence on dedicated infrastructure — combined with the high capital cost of the aircraft themselves and the substantial battery-and-charging electrical demand of operating an eVTOL fleet at hub-airport scale — places eVTOL commercial economics in a fundamentally different category from the conventional commercial aviation infrastructure that has, over more than a century of operational deployment, created the global airport network the broader aviation industry currently operates on.

What 2026 looks like across aviation and airport robotics

In 2026, the aviation and airports robotics category is structurally bifurcated between extremely heavily-regulated, slow-moving, capital-intensive segments (eVTOL aircraft certification, autonomous fixed-wing flight, type-certificated commercial aviation autonomy) and quietly-scaling, less-regulated, operationally focused segments (airport cleaning robotics, baggage handling automation, aircraft inspection drones, counter-UAS detection systems, bird control and wildlife management). The eVTOL category is dominated by the survivors — Joby Aviation (Stage 4 FAA certification, Toyota’s $1 billion-plus invested, Delta partnership, late-2027 commercial-launch target), Archer Aviation (parallel FAA Stage 4 approach, Midnight aircraft, 400+ test flights), Beta Technologies (cargo-focused ALIA 250 development), Wisk Aero (Boeing-backed autonomous variant), Eve Air Mobility (Embraer subsidiary), and the Chinese cohort led by EHang’s CAAC-certificated EH216-S. The category casualties are Lilium and Volocopter, both in active insolvency proceedings as of early 2026. The autonomous fixed-wing aviation category is led by Reliable Robotics (Cessna Caravan autonomy, FedEx integration), Merlin Labs (Caravan and King Air autonomy), and the Joby-Xwing absorption. Counter-UAS is dominated by Axon’s Dedrone subsidiary, DroneShield, Anduril Lattice, Echodyne, and Fortem Technologies. Airport cleaning robotics is dominated by Brain Corp’s BrainOS-powered fleet across 100+ U.S. airports plus Avidbots Neo. Baggage handling is dominated by Vanderlande, Daifuku, and Beumer. Aircraft inspection is dominated by Donecle and Mainblades. Bird control is led by Clear Flight Solutions’ Robird platform.

The structural story across aviation and airport robotics in 2026 is the rigid bifurcation between certification-heavy segments — where multi-billion-dollar capital deployment and decade-long regulatory processes have produced two operationally-bankrupt European leaders alongside a handful of American survivors still pre-revenue — and operationally-quiet segments where conventional industrial robotics deployment has been quietly scaling across the airport ground operations layer for a decade without significant public attention. The Joby November 2025 Stage 4 milestone and the March 2026 FAA powered-lift rule publication represent genuine regulatory progress toward eVTOL commercial operation. The Lilium-Volocopter collapse represents the operational reality that even substantial venture capital deployment, sophisticated engineering, and partial regulatory progress are not sufficient to carry an eVTOL developer to revenue without sustained access to bridge financing at the specific moments certification milestones convert into commercial-readiness milestones.

The robots that are quietly and successfully operating inside the world’s airports in 2026 — the Brain Corp-powered Tennant scrubbers cleaning Hartsfield-Jackson Atlanta’s terminal floors overnight, the Vanderlande baggage handling systems routing 90 million bags annually through Heathrow Terminal 5, the Donecle inspection drones photographing Air France Industries KLM Engineering aircraft in 15-minute pre-flight cycles, the Dedrone radar systems monitoring the perimeters of major U.S. military bases and select hub airports — are not the robots the broader public-attention apparatus directs the conversation toward. The robots receiving public attention are the eVTOL air taxis, which have, in 2026 operational reality, carried zero fare-paying commercial passengers in the United States and Europe and which represent a future deployment scenario whose actual scaling timeline remains substantially uncertain. The aviation and airport robotics category, like the factory robotics category the broader industrial economy depends on, has its operationally important infrastructure layer doing routine work invisibly to the public-attention layer while the high-attention segments fight through the regulatory architecture that aviation, of all industries, applies more rigorously than any other commercial sector.

The 65 years of operational data that the conventional industrial robotics industry generated between Unimate’s 1961 deployment at General Motors and the modern factory floor have, in aviation, been compressed into a substantially shorter window — the first commercial drone operations of any kind required navigating a regulatory architecture that the FAA, ICAO, and EASA built over the course of a century to govern manned aviation safety. The eVTOL industry is now learning, through Lilium and Volocopter’s collapse and through Joby and Archer’s slow grind through five-stage FAA certification, that the regulatory framework that has made commercial aviation the safest mode of mass transportation in human history is not optional for new aircraft categories — it is the operational foundation that commercial scaling requires, in operational contrast to the substantially less heavily regulated autonomous space robotics deployment environment where the absence of human passengers permits substantially faster certification timelines. The airport ground robotics that has quietly scaled across the cleaning, baggage handling, inspection, and security perimeters of the world’s airports has succeeded precisely because it operates inside, rather than against, the regulatory architecture that aviation imposes. The eVTOL industry will scale eventually — Joby, Archer, EHang, and the surviving cohort are working through the certification process that will, on a multi-year timeline, deliver actual commercial operation. The aviation robotics ecosystem in 2026 is, more than any other category in the broader commercial robotics economy, the category where regulatory architecture is the primary variable, technology readiness is the secondary variable, and capital access is the tertiary variable that determines whether any given manufacturer ultimately survives to deliver the operational deployment the venture capital case originally promised.

The Joby November 2025 Stage 4 FAA TIA milestone and the March 27, 2026 FAA powered-lift rule publication are the operational data points that will, eventually, determine whether the surviving U.S. eVTOL cohort produces commercial revenue. The 1,000+ flights cancelled at Gatwick in December 2018 are the operational data point that drove the multi-billion-dollar counter-UAS industry that Axon, DroneShield, Anduril, and Echodyne now compete inside. The 100+ U.S. airports running Brain Corp’s BrainOS cleaning robots are the operational data point that demonstrates what successful airport robotics deployment looks like when the regulatory architecture cooperates rather than obstructs. The aviation industry is, in robotics deployment terms, the most heterogeneous category in the global commercial robotics economy — simultaneously the slowest-moving in the segments that capture public attention and the fastest-scaling in the segments that operate quietly inside the existing regulatory framework. The robots that will ultimately determine whether the eVTOL revolution actually happens are not yet flying commercial passengers. The robots that are already operating successfully inside the world’s airports are the ones cleaning the floors overnight, scanning aircraft for hail damage, sorting checked bags, and detecting unauthorized drones at the airfield perimeter — quietly, profitably, and at deployment scales the eVTOL category will need another half-decade to approach.
Factory and Manufacturing Robots and Drones in 2026: Inside the World’s Largest, Oldest, and Most Operationally Mature Robotics Deployment

In November 2025, a California-based humanoid robotics company called Figure AI announced the official retirement of its Figure 02 humanoid platform after the completion of an 11-month pilot deployment at BMW Manufacturing’s Spartanburg, South Carolina assembly plant. The operational data Figure published with the retirement announcement was the most detailed disclosure ever made of a humanoid robot’s performance in an active commercial factory. Two Figure 02 units, each 170 centimeters tall, 70 kilograms in mass, with a 20-kilogram payload capacity, operated 10-hour shifts Monday through Friday on the BMW X3 body shop line, performing the specific operational task of removing sheet-metal parts from racks and bins and placing them onto welding fixtures with a 5-millimeter tolerance, on an 84-second cycle time (37 seconds for the load alone). The robots accumulated 1,250 hours of runtime, loaded more than 90,000 sheet-metal parts, contributed to the production of more than 30,000 BMW X3 vehicles, walked approximately 1.2 million steps covering an estimated 200 miles inside the plant, and maintained placement accuracy above 99 percent across the deployment. Brett Adcock, Figure’s CEO, accompanied the retirement announcement with photos of the Figure 02 units returning to Figure’s headquarters covered in scratches, scuffs, and industrial grime. The forearm subsystem, by Figure’s own disclosure, was the top hardware failure point. The lessons would, by Figure’s stated plan, be integrated into the next-generation Figure 03 platform launching for production deployment in 2026.

The BMW Spartanburg deployment is, in 2026 operational terms, the most heavily-documented humanoid-robot-in-factory deployment in the commercial history of industrial robotics. It is also, by every available measure of deployed-unit count, an almost negligible fraction of the actual industrial robotics installed base operating inside the world’s factories in 2026. The International Federation of Robotics estimates the global industrial robot installed base passed 4 million units in 2024 — bolted-down articulated arms, SCARA robots, parallel-kinematic delta robots, and collaborative robots operating in continuous production across the automotive, electronics, metals, plastics, food-and-beverage, pharmaceutical, and aerospace manufacturing sectors. The first industrial robot — Unimate, designed by George Devol and Joseph Engelberger — was installed at a General Motors plant in Trenton, New Jersey in 1961. The factory robotics industry has 65 years of operational deployment behind it. The humanoid robot pilots at BMW, Mercedes-Benz Berlin-Marienfelde, Tesla Fremont, GXO Logistics Atlanta, and the growing list of automotive and logistics factory pilots are, in installed-base terms, a few hundred units against an installed base of 4 million conventional industrial robots that have been quietly producing the physical objects of the modern economy since before most of the people designing humanoid robots were born.

The Big Four industrial robot manufacturers

The global industrial robotics market is dominated, by both installed base and annual installations, by four manufacturers: FANUC Corporation (Japan), ABB Group (Switzerland), KUKA AG (Germany), and Yaskawa Electric Corporation (Japan). FANUC, headquartered at the foot of Mount Fuji in Oshino-mura, Yamanashi Prefecture, builds the yellow-painted articulated robots that have become the visual signature of automotive paint shops, electronics assembly lines, and metal-fabrication facilities globally. FANUC’s installed base is approximately 750,000 deployed industrial robots worldwide, with the M-410, R-2000iC, LR Mate, and CRX collaborative robot product lines spanning payload capacities from 4 kilograms (LR Mate) to 2,300 kilograms (M-2000iA, the company’s heaviest articulated arm). FANUC also manufactures the Roboshot injection-molding machines, the Robocut wire EDM machines, and the Robodrill small-machining centers — the broader factory automation product line that has, in operational terms, made FANUC one of the most consistently profitable Japanese industrial conglomerates over the past three decades.

ABB Group, headquartered in Zurich, builds the IRB series of articulated robots and the YuMi dual-arm collaborative robot, with installed-base estimates in the 500,000-600,000 unit range globally. ABB’s industrial automation business operates across the same automotive, electronics, food-and-beverage, and metals manufacturing segments as FANUC, with particular strength in European automotive deployment. KUKA AG, headquartered in Augsburg, Germany, builds the orange-painted KR series of articulated robots that has been operationally synonymous with German automotive manufacturing for decades — KUKA robots populate the assembly lines at Volkswagen, BMW, Mercedes-Benz, and Audi facilities across Europe at deployment volumes no Japanese or American manufacturer approaches. KUKA was acquired by Midea Group — the Chinese consumer appliance conglomerate — in a 2017 transaction that, despite the substantial geopolitical attention it received at the time, has produced relatively continuous operational management since the transaction closed. Yaskawa Electric, the Kitakyushu-based Japanese manufacturer, operates the Motoman robot brand, with the GP, MH, and AR series spanning the standard industrial-robot payload range and an installed base in the 500,000-plus unit range.

The Big Four collectively account for, by industry analyst estimates, approximately 55 to 65 percent of global industrial robot installations in any given year. The remaining 35 to 45 percent is distributed across a long tail of specialist manufacturers — Kawasaki Heavy Industries, Mitsubishi Electric (Melfa series), Denso Corporation (VS series), Stäubli (TX and TS series), Epson Robots (SCARA platforms), Nachi-Fujikoshi, and increasingly the Chinese manufacturers discussed below. The product taxonomy of conventional industrial robots is highly standardized across these manufacturers: six-axis articulated robots for general assembly, SCARA robots for high-speed pick-and-place, delta robots for high-throughput packaging, palletizing robots for warehouse end-of-line operations, and collaborative robots (cobots) for human-robot shared workspace applications. The form factors, control architectures, and operational deployment patterns have, over the past 30 years, converged on a set of standards that the entire factory automation industry operates against.

The cobot category: Universal Robots, Doosan, Techman, and the small-payload collaborative wave

The fastest-growing subcategory within industrial robotics over the past decade has been collaborative robotics — the smaller, force-limited, vision-aware articulated arms designed to operate alongside human workers without traditional safety cages or perimeter fencing. The category-leading manufacturer is Universal Robots, the Danish company founded in 2005 in Odense and acquired by Teradyne (NASDAQ:TER) in 2015 for approximately $285 million. Universal Robots has, as of 2024, deployed more than 75,000 cobots globally across the UR3, UR5, UR10, UR16, and UR20 product lines, with the UR15 platform launching in March 2025 as the company’s most recent product addition. The Universal Robots cobot architecture — a six-axis articulated arm with force-torque sensing at every joint, a polycarbonate enclosure, payloads ranging from 3 kilograms (UR3) to 30 kilograms (UR30), and a unified control architecture that enables relatively rapid task programming compared to traditional industrial robots — has become the dominant operational template for the broader cobot category.

The competing cobot manufacturers include Techman Robot (Taiwan, owned by Quanta Computer since 2018, builder of the TM series cobots with integrated machine vision), Doosan Robotics (South Korea, the M and H series cobots, IPO’d on the Korea Exchange in October 2023), Franka Emika (Munich-based, the Panda cobot platform, restructured under bankruptcy in 2023 and acquired by Cologne-based industrial robotics company Agile Robots SE), AUBO Robotics (Chinese-American joint venture), Productive Robotics (U.S.-based OB7 cobot), and the cobot lines from the Big Four (FANUC CRX, ABB YuMi and GoFa, KUKA LBR iiwa, Yaskawa HC-series). The cobot market in 2026 is estimated at approximately $2.5 billion in annual revenue, with double-digit annual growth rates substantially exceeding the broader industrial robotics market’s mid-single-digit growth.

The Chinese industrial robotics rise: Estun, Inovance, EFORT, and the Made in China 2025 acceleration

The single most operationally consequential shift in factory robotics over the 2020-2026 window has been the rise of Chinese industrial robot manufacturers. China became the world’s largest annual industrial robot market by installations in approximately 2016 and has, by IFR data, accounted for approximately 52 percent of global industrial robot installations in 2024 — more than 290,000 newly-installed robots in China alone against a global total of roughly 560,000 installations. The shift on the demand side was followed by an equally significant shift on the supply side. Estun Automation (Nanjing, Shenzhen-listed under 002747.SZ), Inovance Technology (Shenzhen, listed under 300124.SZ), EFORT Intelligent Equipment (Wuhu, listed under 688165.SH), Siasun Robot & Automation, STEP Electric Corporation, and Han’s Robot have, over the 2018-2026 window, collectively grown from minor domestic players to genuine global competitors. Estun, in particular, has emerged as the largest Chinese industrial robot manufacturer by deployed units, with an installed base in the 100,000-plus range as of 2024 and acquisitions across the European industrial automation supply chain — including the 2017 acquisition of TRIO Motion Technology in the United Kingdom and the 2019 acquisition of German automation specialist Cloos Schweißtechnik.

The structural driver behind the Chinese industrial robotics rise is the Made in China 2025 industrial policy, launched in 2015 by the Chinese State Council, that designated industrial robotics as one of ten priority strategic sectors for domestic capability development. Combined with the broader dual-circulation economic strategy announced in 2020, the policy framework has funneled substantial state-directed investment into Chinese industrial robotics manufacturers, robotic component suppliers (precision reducers, servo motors, controller electronics), and downstream factory automation deployment across Chinese manufacturing. The 2024-2026 acceleration has been driven by the broader decoupling pressures between Chinese manufacturing and Western technology supply chains, with Chinese manufacturers increasingly required by state-directed procurement policies to source domestic industrial automation equipment where viable.

The humanoid robot factory wave: Figure, Tesla Optimus, Apptronik, Agility, and the auto-and-logistics pilot deployment cohort

The humanoid robot wave that began commercial factory pilot deployment over the 2023-2026 window is, in operational terms, the most heavily-financed and most-publicized but smallest-by-deployed-unit-count segment of the broader factory robotics market. The Figure 02 BMW Spartanburg pilot is the most operationally documented example. Tesla‘s Optimus platform has been deployed inside Tesla’s Fremont, California and Austin, Texas vehicle manufacturing facilities for testing and routine task execution, with Elon Musk publicly stating in multiple 2024-2025 earnings calls that Tesla is targeting thousands of Optimus units in internal factory deployment by 2026. Apptronik‘s Apollo platform has been deployed at Mercedes-Benz manufacturing facilities in Berlin-Marienfelde and Kecskemét, Hungary, and inside Jabil electronics-manufacturing operations under the strategic partnership announced in February 2025. Agility Robotics‘ Digit has been deployed at GXO Logistics Spanx fulfillment operations in Atlanta and at additional logistics customer sites. 1X Technologies‘ Neo has been deployed in pilot facilities, with the company having raised more than $100 million from investors including OpenAI. Hexagon Robotics‘ AEON humanoid, unveiled in June 2025, began pilot deployment at BMW’s Leipzig plant in December 2025 as the second humanoid robot deployed within the BMW iFACTORY initiative, alongside the broader Boston Dynamics Spot quadruped fleet that has been operating in BMW and Hyundai factory inspection routines since 2022. Foxconn has, since 2023, publicly disclosed development of humanoid robotics in partnership with NVIDIA’s Project GR00T platform for deployment in its electronics-manufacturing operations, with the underlying foundation-model work increasingly conducted in collaboration with academic robotics research labs at Stanford, MIT, Carnegie Mellon, and UT Austin.

The structural observation about the humanoid factory wave in 2026 is that the total deployed unit count across all manufacturers globally is, by available public disclosure, in the low thousands — roughly 0.05 to 0.1 percent of the broader industrial-robot installed base. The pilots are operationally important. The Figure 02 BMW deployment has generated more public-facing data about humanoid factory performance than any prior deployment. The Tesla Optimus internal deployments — though Tesla has disclosed less specific operational data than Figure has — have, by Musk’s public claims, achieved meaningful internal factory utility. But the bolted-down FANUC, ABB, KUKA, and Yaskawa industrial robots that have populated the world’s factories for 60 years continue to outnumber the humanoid platforms by approximately 1,000 to 1 in deployed-unit terms, and continue to perform the bulk of the actual manufacturing work in the global economy in 2026.

Robot density: South Korea, Singapore, Germany, Japan, and the international competitiveness story

The most useful single statistic for understanding the international competitive dynamics of factory automation is robot density — the number of operational industrial robots per 10,000 manufacturing workers in a given economy. IFR data for 2022-2023 placed South Korea at approximately 1,012 robots per 10,000 manufacturing workers — the highest robot density in any major economy in the world by a significant margin. Singapore was second at approximately 770. Germany was third at approximately 415. Japan was fourth at approximately 397. China had climbed to fifth place at approximately 322 robots per 10,000 manufacturing workers — a substantial increase from sub-100 a decade earlier. The United States was sixth at approximately 285, with Sweden, Denmark, Hong Kong, and Taiwan rounding out the top ten. The implication for U.S. manufacturing competitiveness is direct: South Korea operates approximately 3.5 times more industrial robots per manufacturing worker than the U.S. does, and the gap has been widening since approximately 2018 rather than narrowing.

The structural driver behind the South Korean robot-density lead is the heavy concentration of South Korean manufacturing in two sectors — automotive (Hyundai, Kia, KG Mobility) and electronics (Samsung, LG, SK Hynix) — both of which are extremely high-automation industries by global standards, and both of which have been actively automating since the 1990s under coordinated industrial policy. The structural driver behind the Singapore robot-density figure is the electronics manufacturing concentration in the Singaporean economy combined with active state-led automation incentives. The structural driver behind the German robot-density is the legacy of German automotive manufacturing’s longstanding automation leadership and the broader Mittelstand mechanical-engineering ecosystem. The structural driver behind the U.S. relative lag is harder to summarize cleanly — the U.S. manufacturing sector is more heterogeneous (broader range of industries), the labor cost gap between manual labor and automation has been smaller for most of the past 30 years than in higher-cost economies, and the historical U.S. manufacturing offshoring wave to Mexico, China, and Southeast Asia reduced the demand for domestic factory automation through the 2000s and 2010s.

The reshoring wave and the CHIPS Act / IRA / IIJA buildout context

The single largest demand-side accelerator for U.S. factory robotics in the 2024-2026 window has been the convergence of three federal industrial-policy initiatives: the CHIPS and Science Act (signed August 2022, authorizing approximately $52 billion in semiconductor manufacturing incentives), the Inflation Reduction Act (August 2022, approximately $369 billion in clean energy spending including electric vehicle and battery manufacturing incentives), and the Infrastructure Investment and Jobs Act (November 2021, $1.2 trillion in infrastructure spending). The CHIPS Act has driven major semiconductor manufacturing facility construction at TSMC Arizona (Phoenix), Intel Ohio (New Albany), Samsung Texas (Taylor), Micron New York (Syracuse), and GlobalFoundries New York (Malta). The IRA has driven major battery and EV manufacturing facility buildouts at Tesla Gigafactory Nevada (expansion), Tesla Gigafactory Texas (Austin), Hyundai Metaplant (Bryan County, Georgia), Ford BlueOval City (Tennessee), Volkswagen Scout Motors (South Carolina), and LG Energy Solution, SK Innovation, Panasonic, and CATL battery manufacturing investments across multiple U.S. states. Each of these new facilities represents tens of thousands of square feet of greenfield factory floor space requiring industrial robotics deployment from initial buildout, and each represents capital deployment that conventional manufacturing-equipment depreciation cycles would otherwise have spread across decades.

The structural reshoring trend has, by every available measure, been the most consequential single demand driver for U.S. factory automation since the 1990s. The factories being built are being built with substantially higher automation densities than the U.S. manufacturing facilities they are notionally replacing, in part because the labor cost equation no longer supports manual-labor-intensive operations at U.S. wage levels and in part because the semiconductor and battery manufacturing processes being deployed are inherently more automation-dependent than the consumer electronics and automotive operations that previous waves of U.S. manufacturing offshored.

The factory drone category: Verity, Pinc Solutions, and the indoor inventory inspection niche

The drone category in factory operations is, in operational terms, much smaller than the industrial-robot category, but it occupies a specific niche around indoor inventory inspection and asset surveillance. Verity AG, the Zurich-based industrial drone company, builds fully-autonomous indoor drones that operate inside warehouses and distribution centers, scanning RFID-tagged inventory pallets, capturing visual documentation of stock positions, and feeding data into warehouse-management systems. Verity has deployed across Nestlé, Maersk, DSV, and Geodis warehouse operations. Pinc Solutions operates a competing indoor inventory drone platform deployed at Ralph Lauren, Lego, and Bridgestone distribution facilities. Eyesee (a subsidiary of Hardis Group, France) operates the Eyesee indoor warehouse inventory drone. The indoor warehouse drone category, while smaller in revenue than the broader industrial-robot category, has demonstrated the operational use case for autonomous aerial robotics in structured indoor environments where the outdoor drone navigation challenges do not apply.

The outdoor factory drone category — perimeter security, smokestack and refinery inspection, solar array inspection, large facility surveying — is dominated by the same drone manufacturers serving construction and infrastructure inspection markets: DJI (Phantom 4 RTK, Matrice 350 RTK, Mavic 3 Enterprise), Skydio, Parrot Anafi USA, and Flyability‘s Elios confined-space inspection drone, which operates inside boilers, storage tanks, and other enclosed industrial spaces.

What 2026 looks like across factory and manufacturing robotics

In 2026, the factory robotics category is structurally dominated by the conventional industrial robot installed base — approximately 4 million deployed units globally, growing by 500,000-plus annual installations, dominated by FANUC, ABB, KUKA, and Yaskawa with the long tail of specialist manufacturers and the rapidly-growing Chinese manufacturers (Estun, Inovance, EFORT) accounting for the balance. The cobot category, dominated by Universal Robots with Techman, Doosan, and the Big Four’s cobot lines competing, continues to be the fastest-growing subcategory at approximately $2.5 billion in annual revenue. The humanoid factory wave — Figure (post-02 retirement, transitioning to Figure 03), Tesla Optimus, Apptronik Apollo (Mercedes-Benz, Jabil), Agility Digit (GXO, Amazon), 1X Neo, Hexagon AEON (BMW Leipzig), and the Foxconn-NVIDIA humanoid manufacturing initiative — operates at deployed-unit volumes in the low thousands against the four-million-unit conventional installed base, with the Figure 02 BMW Spartanburg deployment standing as the most operationally documented humanoid-in-factory deployment in commercial history. South Korea operates at 1,012 robots per 10,000 manufacturing workers; the U.S. operates at 285. The CHIPS Act, IRA, and IIJA federal industrial policy is driving the largest U.S. factory buildout in three decades, with TSMC Arizona, Intel Ohio, Samsung Texas, and the broader EV-and-battery manufacturing investment wave creating the demand environment for accelerated factory automation deployment.

The structural story across factory robotics in 2026 is that the category is, simultaneously, the most operationally mature and the most actively disrupted of any robotics deployment domain. The bolted-down industrial robot has 65 years of operational deployment behind it — six decades that no other robotics category approaches. The 4 million installed units perform the bulk of the actual manufacturing work in the global economy and will continue to do so for the operational lifetime of the equipment currently deployed. But the category is also being actively disrupted on multiple vectors simultaneously: Chinese manufacturers competing with the historical Big Four on cost and increasingly on capability, cobot manufacturers expanding the addressable market into smaller manufacturers that conventional industrial robots could not serve, humanoid robot manufacturers piloting platforms that — if the operational reliability projected by Figure, Tesla, Apptronik, Agility, and 1X actually materializes at scale — could expand the addressable factory-automation market by an order of magnitude over the 2026-2035 window. The category is dominated by mature platforms doing routine work, layered over by a small number of high-attention-receiving experimental platforms that may or may not eventually justify the venture capital and corporate-strategic investment they have received.

The Figure 02 BMW deployment is the operational data point that defines what the answer might look like. Eleven months. 1,250 hours. 90,000 sheet-metal parts. 30,000 BMW X3 vehicles. 99 percent placement accuracy. A forearm subsystem that emerged as the top hardware failure point — and a Figure 03 platform launching in 2026 that will, by Figure’s stated plan, address the specific hardware reliability lessons learned at Spartanburg. The traditional six-axis FANUC welding robot down the line that received the sheet-metal parts the Figure 02 robots loaded did not generate a press release. The traditional robot has been doing that exact task in some configuration since approximately 1985. The traditional robot is the deployed industrial economy. The humanoid platform is the deployment experiment that, depending on how the Figure 03 / Optimus / Apollo / Digit / AEON / Neo cohort performs over the 2026-2030 window, could either become the next mature deployment template or could remain a high-visibility experimental category that the conventional industrial-robot installed base ultimately absorbs without fundamental architectural change.

The data that will resolve that question over the next five years is being generated, in 2026, inside the same global factory installed base that has been quietly producing the physical objects of the modern economy for six decades. The robots that move the global trade flows, patrol oil-and-gas facilities, deliver hospital prescriptions to patient homes, retrofit excavators into autonomous solar pile drivers, respond to wildfires and structural collapses, scout planetary surfaces beyond Earth, count penguins in Antarctica, and throw 100-mph cutters in MLB clubhouses all derive, in mechanical engineering, control architecture, and operational deployment terms, from the bolted-down industrial robot that George Devol and Joseph Engelberger installed at the General Motors Trenton plant in 1961. The factory is the parent industry. Everything else is a derivative deployment of the operational principles that the factory automation industry has been refining since the Eisenhower administration. The robots that work at scale in 2026 — anywhere in the economy, in any application — work because the conventional industrial-robot industry figured out, six decades ago, that automation is not about replacing humans wholesale but about deploying specialized machines for specific repetitive tasks under operational constraints that the broader industrial supply chain can actually sustain. The Figure 02 BMW pilot is, in operational terms, the same kind of deployment experiment that General Motors ran with Unimate in 1961. The result, after sixty-five years of cumulative learning, is the 4-million-unit global installed base that quietly produces almost everything else.

The next sixty-five years will be either an extension of that operational logic into humanoid-robot territory or a continuation of the bolted-down articulated-arm dominance that has, on the available evidence, been the most successful single deployment template in the history of industrial automation. Which of those two outcomes materializes depends on a small number of specific operational variables — humanoid hardware reliability at scale, the training of the next generation of robotics engineers, the comparative cost trajectories of humanoid versus conventional platforms — that are being actively worked on inside Figure, Tesla, Apptronik, Agility, 1X, FANUC, ABB, KUKA, Yaskawa, Estun, and the broader factory robotics industry in 2026. The answer is not yet known. The deployment data being generated in the meantime, including the Figure 02 / BMW Spartanburg pilot, is what will eventually determine which template wins.
Construction Robots and Drones in 2026: The Industry Where Automation Took Half a Century Longer Than Everyone Else

In September 2025, the utility-scale solar construction subsidiary of Quanta Services — a company called Blattner that operates as one of the largest engineering-procurement-construction (EPC) contractors in U.S. renewable energy infrastructure — announced it was deploying dozens of autonomous solar pile-driving robots built by a San Francisco-based construction-robotics startup called Built Robotics on the company’s nationwide solar installation projects. The robots in question are not new platforms purpose-built for autonomy. They are conventional hydraulic excavators — the same Caterpillar, Komatsu, Volvo, and Hitachi excavators that have been operating on construction sites since the mid-twentieth century — retrofitted with Built Robotics’ Exosystem, an aftermarket autonomy upgrade kit that converts a manually-operated excavator into a fully-autonomous robot in approximately four hours of installation time and that, critically, remains fully reversible. The Exosystem sits below the excavator’s boom mobilization height, so the machine remains transportable. The system includes six 360-degree onboard cameras, RTK GPS positioning accurate to centimeters, IMU-based kinematic software, an all-weather ruggedized enclosure, and a liquid-cooled embedded computing platform. The robot operates 24 hours a day on solar pile-driving projects, requires only periodic resupply and refueling, and has demonstrated production rates of approximately 2.5 times the equivalent human-operated baseline. Built Robotics CEO Noah Ready-Campbell — the former Google engineer who founded the company in 2016 and who has, over the intervening decade, become one of the most identifiable figures in U.S. construction robotics — publicly framed the deployment thesis around 24/7 operation enabling project schedule acceleration in a way that conventional construction crews structurally cannot.

The Built Robotics-Blattner partnership is the cleanest single illustration of the structural argument that has, over the past decade, finally begun to unlock construction robotics as a commercial category: construction does not automate the way warehouse logistics, factory manufacturing, or hospital operations automate. Construction sites are heterogeneous by definition — every project has different terrain, different blueprints, different weather, different crews, different supply chains, different regulatory environments, different existing infrastructure to work around. The general-purpose humanoid robot that operates inside a structured Mercedes-Benz factory floor or an Amazon fulfillment center cannot, in any practical 2026 sense, walk onto a residential construction site and frame a house. The category that has succeeded in construction is the category that picked a single repetitive task — pile driving, drywall installation, layout marking, brick laying, demolition, site survey — automated that one task at scale, and let humans handle everything else. The successful construction-robotics platforms are not general-purpose. They are surgically specialized.

Why construction is the last major industry to automate

The fundamental productivity statistic that defines the construction-robotics market opportunity is the McKinsey Global Institute analysis showing that U.S. construction productivity has been approximately flat over the past 50 years, while manufacturing productivity has grown by approximately seven times over the same period. Construction is, by every available labor-productivity measure, the largest U.S. industry that has not meaningfully been transformed by automation. The structural reasons are well-documented. Construction projects are bespoke. Construction sites are outdoor, weather-exposed, and physically chaotic. Construction crews are heterogeneous — the same project can involve dozens of subcontractors from different trades, each operating on different schedules and with different equipment. The regulatory environment is fragmented across federal, state, and municipal jurisdictions. The supply chain is project-specific. The skilled labor pool is, in 2024-2026 terms, severely undersupplied — the Associated General Contractors of America estimated U.S. construction needed approximately 500,000 additional workers in 2024 above existing employment to meet demand, with the underlying skilled-trades training pipeline producing replacement workers at substantially lower rates than the construction-industry retirement and turnover curve requires, and with the underlying labor shortage projected to persist through the late 2020s. These structural conditions are simultaneously the reason construction has not been automated historically and the reason automation has, in the 2020s, finally become economically viable. The labor cost is rising fast enough, and the project-volume demand is large enough, that the return-on-investment math has shifted in favor of specialized robotic platforms in a way it has not previously supported.

The single largest demand-side driver of construction-robotics investment in the 2020s has been federal infrastructure spending. The Infrastructure Investment and Jobs Act (IIJA) signed in November 2021 authorized approximately $1.2 trillion in federal infrastructure spending across roads, bridges, public transit, broadband, water systems, and electric grid upgrades. The Inflation Reduction Act (IRA) signed in August 2022 authorized approximately $369 billion in clean energy and climate-related spending, including the solar tax credits and renewable-energy investment incentives that have driven the utility-scale solar construction boom Built Robotics is now servicing. These two pieces of legislation, in combined dollar volume, represent the largest peacetime federal infrastructure capital deployment in U.S. history, and they have created the multi-year construction-demand environment that has made specialized robotic platforms economically defensible at unit-deployment scale.

The 3D-printed residential construction story: ICON, Wolf Ranch, and the Lennar deployment

The most operationally consequential 3D-printing construction company in the United States is ICON, an Austin, Texas-based construction technology company that operates the Vulcan robotic construction system. The Vulcan printer is, in physical terms, an approximately 46.6-foot-wide by 15.6-foot-tall robotic gantry that extrudes a proprietary cement-based material called Lavacrete through a nozzle in successive horizontal layers, building the walls of a single-family home in approximately three weeks of printing time per unit, with the foundation and metal roof installed using conventional construction methods. ICON’s flagship deployment is the Wolf Ranch community in Georgetown, Texas — a 100-home master-planned development north of Austin, built in partnership with national homebuilder Lennar Corporation (NYSE:LEN) and co-designed by Danish architectural firm BIG-Bjarke Ingels Group. The Wolf Ranch homes range from 1,500 to 2,100 square feet, with three to four bedrooms, and were priced starting in the mid-$400,000s at the project’s initial sales launch in 2023. The development is part of Hillwood Communities, a Perot Company. As of August 2024, more than 80 percent of the Genesis Collection homes had sold, with the first homeowners moving in beginning September 2023. The Wolf Ranch project is, in 2026 operational terms, the largest 3D-printed residential community ever completed anywhere in the world.

ICON’s broader portfolio extends beyond Wolf Ranch. The company has partnered with the Texas Military Department on 3D-printed military barracks construction. The company built the first 3D-printed homes for Habitat for Humanity in Williamson County, Texas. ICON has additional 3D-printing deployments in El Cosmico, the BIG-co-designed glamping resort expansion in Marfa, Texas, with home prices reaching into the seven figures for the larger custom units. The company’s Vulcan printer is a multi-million-dollar piece of capital equipment that requires specialized operators, customized proprietary materials, and ongoing engineering support. The 3D-printing residential construction category in 2026 is, in industry-wide terms, still small — ICON, Apis Cor, COBOD International (the Danish manufacturer that supplies Vulcan-style construction printers to international markets), and a handful of smaller specialist competitors collectively account for low-four-figure units of completed 3D-printed housing globally — but the category is growing at the highest rate of any subcategory in residential construction technology.

The autonomous heavy equipment category: Caterpillar Command, Komatsu Smart Construction, and the Built Robotics retrofit thesis

The largest single category of construction robotics by deployed unit count is autonomous heavy equipment, dominated by the major incumbent manufacturers — Caterpillar, Komatsu, Volvo Construction Equipment, Hitachi Construction Machinery, and Chinese manufacturer Sany. Caterpillar’s Command for hauling autonomous truck system has been operationally deployed across multiple large-scale mining operations since 2013, with more than 500 autonomous haul trucks operating across BHP, Rio Tinto, Fortescue, and Suncor mining sites globally as of 2024. Komatsu operates the Smart Construction platform, which integrates autonomous bulldozer operation, drone-based site survey, and BIM-driven excavation planning into a single integrated workflow. Volvo CE has demonstrated the TARA autonomous hauler. The autonomous heavy equipment category, when measured by total deployed-unit count, dwarfs every other construction-robotics category — but the deployed units are heavily concentrated in mining, aggregates, and large-scale resource extraction rather than in conventional building construction, where site heterogeneity makes autonomous-equipment deployment substantially harder.

The Built Robotics thesis — retrofit aftermarket autonomy onto existing fleets of conventional excavators rather than selling purpose-built autonomous platforms — represents a different commercial bet. The Exosystem retrofit kit can be installed on mid-size excavators from any of the major manufacturers, the installation is reversible, and the business model bills as a combined monthly rental fee plus hourly operation wage rather than a large upfront capital purchase. The company’s pivot from general construction trenching to solar farm pile driving, announced in 2023 and consummated through the Blattner partnership in 2025, reflects the structural lesson that has emerged across construction robotics: the path to commercial scale runs through specialized, repetitive, high-volume applications rather than through general-purpose automation. Built Robotics’ RPD 35 autonomous pile-driving platform is the operational expression of this thesis. The platform was granted a U.S. patent for the autonomous pile-driving system in February 2025. The deployment focus is on U.S. and Australian solar markets through 2026.

The specialized indoor-construction robots: Dusty Robotics, Canvas, Hadrian X, and Hilti Jaibot

The indoor-construction specialty-robot category includes a growing number of platforms each focused on a single repetitive task. Dusty Robotics, the Mountain View-based construction-robotics company founded in 2018 by Tessa Lau and Philipp Herzig, builds the FieldPrinter — a small, wheeled, ground-printing robot that automatically marks construction layouts on concrete slabs from BIM model data. The platform replaces the manual chalk-line and tape-measure layout process that has, for decades, been one of the most labor-intensive and error-prone steps in commercial construction, with FieldPrinter deployments documented across major U.S. general contractors including DPR Construction and Suffolk. Canvas, the San Francisco-based drywall-finishing robotics company founded in 2017, builds an autonomous platform that applies and finishes drywall joint compound — taping, mudding, and sanding — using a robotic arm mounted on a mobile base, with the platform’s first commercial deployments concentrated in Bay Area commercial construction projects. Fastbrick Robotics, the Australian company that builds the Hadrian X automated brick-laying robot, operates a truck-mounted articulated boom that places bricks at a documented rate of approximately 200 bricks per hour, in continuous operation, with the first commercial home deployments completed in Western Australia and the platform being expanded into the U.S. and Mexican markets through partnerships with Wienerberger and other major brick producers. Hilti, the Liechtenstein-based construction tool manufacturer, operates the Jaibot — a semi-autonomous overhead drilling robot designed for the high-volume drilling required in mechanical, electrical, and plumbing (MEP) ceiling installations in commercial construction, marketed as a way to reduce the repetitive overhead labor that contributes disproportionately to construction-trade musculoskeletal injuries.

The demolition robot category: Brokk and Husqvarna

The demolition robotics subcategory operates with a different operational logic than the rest of construction robotics. Demolition robots are remote-operated rather than autonomous. They are designed primarily to remove humans from environments where structural collapse, asbestos exposure, or radiological contamination would make manual demolition unacceptably dangerous. The category leader is Brokk, the Skellefteå, Sweden-based manufacturer that has, since 1976, produced compact electric-and-hydraulic demolition robots ranging from the Brokk 70 (170 kilograms, designed for tight indoor spaces) through the Brokk 900 (10,500 kilograms, designed for large-scale industrial demolition). Brokk robots have been deployed in nuclear decommissioning at Sellafield in the United Kingdom, at Fukushima Daiichi in the post-2011 reactor stabilization operation, and across major infrastructure renovation projects globally. Husqvarna, the Swedish power equipment manufacturer, builds the competing DXR demolition robot line. The demolition robot category is, in commercial terms, smaller than autonomous-heavy-equipment or specialty-indoor-robot categories — but the platforms operate in environments where the alternative to robotic deployment is either prohibitive worker risk or non-completion of the project.

The site-monitoring robot category: Boston Dynamics Spot at Skanska, Suffolk, and the general-contractor wave

The site-monitoring robotics subcategory has, since approximately 2020, been dominated by Boston Dynamics’ Spot quadruped platform deployed by major general contractors for daily site documentation, BIM-comparison verification, safety inspection, and progress tracking. Spot deployments at major U.S. and international general contractors include Skanska, Suffolk Construction, Brasfield & Gorrie, Pomerleau in Canada, Foster + Partners‘s construction documentation operations, and the Pomerleau-Built Robotics consortium that has piloted combined autonomous-equipment-plus-site-monitoring workflows. Spot’s site-monitoring deployment typically involves a robot equipped with a 360-degree camera and laser scanner walking a pre-programmed route through an active construction site at regular intervals — typically daily — capturing high-resolution imagery and point-cloud data that is then processed against the project’s BIM model to identify construction deviations, safety violations, and progress milestones. The structural value proposition is data continuity: a human inspector visits a site weekly, while a Spot deployment generates daily documentation, producing a temporal density of project-state data that no human inspection process can match.

The reality-capture software category that processes Spot’s output and competing aerial drone imagery is dominated by OpenSpace, HoloBuilder (acquired by FARO Technologies in 2021), DroneDeploy, and Procore Technologies (NYSE:PCOR). These platforms transform raw drone, robot, and 360-camera imagery into spatially-indexed, BIM-aligned, time-series construction documentation that has become standard practice across major general contractors in the United States.

The drone surveying and aerial photogrammetry category

The construction-site drone category, separate from the indoor-robot category, is dominated by DJI — the Shenzhen-based drone manufacturer that has, despite the ongoing U.S. federal procurement restrictions and the broader scrutiny of Chinese commercial drone technology, continued to operate as the de facto standard for commercial construction site surveying. The DJI Phantom 4 RTK and Matrice 350 RTK platforms operate across U.S. commercial construction sites in volumes that no other manufacturer approaches, with the platforms typically deployed for weekly photogrammetric site surveys, monthly volumetric calculations of aggregate stockpiles, quarterly progress documentation, and incident-specific aerial documentation when safety or quality issues require it. Skydio, the San Mateo-based autonomous-drone manufacturer that has positioned itself as the U.S.-government-approved alternative to DJI, has captured share in federally-funded construction projects and infrastructure inspection deployments. Parrot Anafi USA, the federal-compliant drone built by French manufacturer Parrot, operates in the same federal-procurement segment. Wingtra, the Swiss fixed-wing survey drone manufacturer, operates in the larger-area aerial photogrammetry segment where multirotor drone endurance becomes constraining. AgEagle and Sentera operate adjacent platforms primarily marketed for agricultural and land-management surveying but used in some construction-site applications.

The Katerra collapse and the prefab modular construction cautionary tale

The construction-technology category is not without its operational casualties, and the largest single failure in the recent history of construction robotics and prefabrication is the Katerra collapse. Katerra was founded in 2015 by Michael Marks (the former Flextronics CEO), Fritz Wolff, and Jim Davidson, with the thesis that construction could be transformed by applying manufacturing-industry vertical-integration logic to residential and commercial building production. The company raised more than $2 billion in venture capital, including a $865 million round led by SoftBank Vision Fund in 2018. Katerra acquired multiple architectural firms, engineering firms, and prefabrication factories. The company filed for Chapter 11 bankruptcy in June 2021 after, by available reporting, burning through the bulk of its capital on factory buildouts that never achieved sustainable unit-economics. The Katerra collapse is the clearest single counterexample to the thesis that construction can be straightforwardly automated by importing factory-manufacturing logic into the construction process. The successful 2020s construction-robotics companies — Built Robotics, Dusty, Canvas, ICON, Hadrian X — have all taken a different operational approach. They have not tried to vertically integrate the construction industry. They have taken individual repetitive tasks and automated them in isolation, leaving the rest of the construction value chain unchanged.

What 2026 looks like across construction robotics and drones

In 2026, the construction-robotics category is structurally distributed across a small number of operationally dominant platforms in each subcategory. Autonomous heavy equipment is dominated by the major incumbent manufacturers (Caterpillar Command, Komatsu Smart Construction, Volvo CE TARA) operating primarily in mining and aggregates, with Built Robotics’ Exosystem retrofit platform operating in the specialized solar pile-driving application. 3D-printed residential construction is dominated by ICON’s Vulcan platform, with the Wolf Ranch deployment as the operational proof point and Apis Cor, COBOD, and smaller specialists competing in the broader global market. Indoor specialty robots are dominated by Dusty FieldPrinter (BIM-driven layout marking), Canvas (drywall finishing), Hadrian X (brick laying), and Hilti Jaibot (overhead MEP drilling). Demolition is dominated by Brokk and Husqvarna DXR. Site monitoring is dominated by Boston Dynamics Spot at major general contractors, with OpenSpace, HoloBuilder (FARO), and DroneDeploy as the reality-capture software layer. Aerial surveying is dominated by DJI Phantom 4 RTK and Matrice 350 RTK, with Skydio and Parrot capturing the federal-procurement-restricted segment. The underlying market is, by industry analyst estimates, approximately $4-6 billion annually in 2026 across all construction-robotics subcategories combined, with double-digit annual growth driven by the IIJA and IRA infrastructure spending wave and the persistent construction labor shortage.

The structural story across construction robotics in 2026 is the opposite of the structural story in factory humanoid robotics. The factory humanoid thesis — most aggressively expressed by Tesla Optimus, Figure 02, Apptronik Apollo, and Agility Digit — is that a general-purpose bipedal platform will eventually be flexible enough to perform any task in a structured factory environment, replacing human labor on a task-substitution basis. The construction robotics thesis is the inverse. The successful construction-robotics platforms have all converged on the observation that construction sites are too heterogeneous, too weather-exposed, too physically chaotic, and too project-specific for a general-purpose platform to operate reliably. The path to commercial success runs through hyper-specialization. Build a robot that does pile driving. Build a robot that does drywall. Build a robot that does brick laying. Build a robot that does layout printing. Do not build a robot that does construction generally, because construction generally is the most heterogeneous physical operation in the modern economy and no single platform is going to do all of it.

The deployed-robot fleets that exist in 2026 reflect this convergence. There is no humanoid robot operating on a U.S. construction site in any commercially-significant volume. The Tesla Optimus, Figure, and Apptronik platforms that have accumulated thousands of deployment hours in Mercedes-Benz, BMW, and GXO Logistics facilities have, as of public disclosure, zero deployment hours on conventional construction sites. The Boston Dynamics Atlas humanoid that has accumulated extensive public demonstration footage of parkour and gymnastic movements has not, in any documented commercial sense, been deployed for construction work. The construction-robotics platforms that operate at meaningful commercial scale are wheeled, tracked, or articulated industrial machines that have been retrofitted or purpose-built for a single specialized task. The form factor that has succeeded in this category is, structurally and operationally, the form factor that pre-existed humanoid robotics — the heavy equipment chassis, the gantry printer, the wheeled mobile base, the truck-mounted articulated boom — augmented with the autonomy, computer vision, and embedded computing capability that has emerged across the broader industrial robotics economy in the 2020s.

The question that defines the next decade of construction robotics is whether this hyper-specialized convergence will continue, or whether the general-purpose humanoid platforms will eventually become reliable enough, mobile enough, and weather-resistant enough to operate on construction sites at all. The available evidence in 2026 is that the hyper-specialized convergence will continue. Construction sites are not Mercedes factories. They are not Amazon warehouses. They are not hospital corridors or fulfillment centers or any other operationally-structured environment where a humanoid platform can be trained to perform routine tasks. Construction sites are improvised, weather-exposed, multi-trade environments where the only operating logic that has, over the past decade of attempted automation, actually worked is the logic of automating one repetitive task at a time and leaving everything else to the human workforce, in direct contrast to the generalist deployment thesis driving the commercial humanoid robotics industry.

The Built Robotics-Blattner solar pile-driving partnership is, in 2026 operational terms, the cleanest illustration of what successful construction robotics looks like. A specialized robotic platform automating a single repetitive task — driving steel piles into the ground for solar array foundations — at a 2.5x productivity multiplier over manual operation, 24 hours a day, across an enormous addressable market created by the federal renewable-energy spending wave. The robot doesn’t try to do anything else. It doesn’t have to. The construction industry, after fifty years of frustrated automation attempts, has finally figured out that the way to put robots on construction sites is to put them on construction sites one task at a time. The robots that work are the robots that do less, more reliably, in the specific operational niche where their physical constraints align with the project’s repetitive labor demands. The pipeline of federal infrastructure spending and the persistent construction labor shortage have, between them, created the demand environment that finally makes specialized construction robotics economically defensible. The 2026 operational reality is that the construction industry is being automated, but not the way the general-purpose humanoid evangelists predicted. It is being automated the way the heavy-equipment industry was always going to automate — task by task, machine by machine, retrofit by retrofit, with humans doing what humans do best and robots doing what robots do best, on construction sites that have, after a half-century of resistance, finally become economically viable to put robots on.
Humanoid Robots and Drones in Space in 2026: Stations, Satellites, Lunar Landers and Other Worlds

In February 2025, an Austin-based humanoid robotics company called Apptronik closed a $403 million Series A funding round at a reported $5 billion valuation, with Mercedes-Benz, Google DeepMind, B Capital, Capital Factory, Japan Post Capital, and ARK Invest among the named investors. The company’s flagship humanoid robot — a 5-foot-8, 55-pound-payload-capacity bipedal platform with a sleek white finish that distinguishes it visually from the dark or metallic platforms built by Tesla, Figure, Boston Dynamics, and Unitree — is named Apollo. The naming is not incidental. Apptronik was founded in 2016 by Jeffrey Cardenas, Nicholas Paine, and Luis Sentis, all alumni of the Human Centered Robotics Laboratory at the University of Texas at Austin, where key team members worked on NASA’s Valkyrie humanoid robot program — the 6-foot-2, 300-pound disaster-response and space-exploration humanoid that NASA’s Johnson Space Center developed in the mid-2010s and that became the most ambitious humanoid-in-space platform the U.S. space program has ever publicly funded. The Apollo robot is, in mechanical-engineering pedigree terms, a direct descendant of NASA’s most serious attempt to build a humanoid that could operate alongside astronauts in spacecraft and on planetary surfaces.

The structural irony of Apptronik’s Apollo — and the central paradox that defines the intersection of humanoid robotics and space exploration in 2026 — is that despite the NASA genealogy and the deliberate naming, Apollo is not going to space. Apollo is being deployed in Mercedes-Benz manufacturing plants in Berlin-Marienfelde and Kecskemét, Hungary, in Jabil electronics-manufacturing facilities under a February 2025 strategic partnership, and in GXO Logistics distribution centers under a 2024 multi-phase R&D agreement. The Mercedes deployment involves moving components and performing quality checks at the company’s Digital Factory Campus. The Jabil deployment is structured around “robots building robots” — Apollo units being used to manufacture more Apollo units inside Jabil’s own electronics plants. The robot’s warehouse and factory-floor deployment is, in commercial terms, an enormous business. The robot’s deployment in actual space — on the International Space Station, on lunar surface missions, on Mars — is, as of the 2026 product roadmap publicly disclosed by Apptronik, zero units.

This is the structural pattern that defines the entire humanoid-robots-in-space category in 2026. The humanoid platforms with the strongest NASA pedigree are being commercialized for terrestrial factory work. The robots actually doing useful work in space are, almost without exception, not humanoid — they are free-flying cubes, rotorcraft, wheeled rovers, dedicated robotic arms, and increasingly autonomous satellite buses. The reasons are not mysterious. The space environment is the worst possible operational context for bipedal locomotion: microgravity makes the entire concept of “walking” meaningless on a space station, vacuum demands specialized seals and lubricants that ground-based platforms do not use, radiation degrades semiconductor electronics rapidly enough that the same chips that work for a decade in a Tesla factory will fail within months in low Earth orbit, and every kilogram launched to orbit costs between $1,500 and $10,000 depending on the launch vehicle. A 175-pound humanoid robot like Apollo costs, in launch terms alone, between $260,000 and $1.75 million just to get to the ISS, before the cost of the robot itself and before any consideration of the maintenance windows, spare parts inventory, and engineering support that a complex bipedal platform requires. The space-robotics industry has, over six decades of practical experience, converged on form factors that have nothing to do with the human body and everything to do with the operational constraints of the destination.

The humanoid-in-space history: Robonaut, Skybot FEDOR, and the failed promise

The U.S. side of the humanoid-in-space history is dominated by Robonaut, NASA’s joint program with General Motors that produced Robonaut 2 (R2) — a humanoid upper-body torso with two seven-degree-of-freedom arms, dexterous hands, and a head-mounted sensor suite that was launched to the ISS aboard Space Shuttle Discovery’s STS-133 mission in February 2011. R2 was the first humanoid robot in space. It was, by every measure of the program’s stated objectives, a disappointment. R2 was designed to perform routine maintenance tasks on the ISS interior, freeing astronaut crew time for higher-value scientific work. In practice, R2 spent most of its time on the ISS either powered down or being repaired. A 2014 leg-attachment upgrade — designed to give R2 mobility within the station — never functioned correctly. The robot developed an intermittent electrical fault in 2015 that the crew could not reliably diagnose in microgravity, and in 2018 NASA returned R2 to Earth aboard a SpaceX Dragon resupply capsule for ground-based repair. The robot has not returned to space. NASA’s Valkyrie (also called R5), the ground-based humanoid developed at Johnson Space Center in 2013 for the DARPA Robotics Challenge and subsequently positioned as a candidate for Mars surface missions, has never flown. Valkyrie units exist at the University of Texas at Austin (where Apptronik’s founders worked on the platform), at MIT, at Northeastern, and at NASA’s Johnson Space Center as a research platform. None of them have been to space, and NASA has not publicly committed to a flight mission for the platform.

The Russian side of the humanoid-in-space history is dominated by Skybot F-850, also known as FEDOR (Final Experimental Demonstration Object Research), an anthropomorphic robot built by Android Technics and the Foundation for Advanced Research Projects in the Defense Industry that was launched to the ISS aboard a Soyuz MS-14 mission in August 2019 as the sole cosmonaut on an uncrewed test flight. FEDOR’s stated mission was to demonstrate the capability for a humanoid robot to perform spacecraft operations in microgravity. The robot’s actual achievements on the ISS were limited. FEDOR was photographed, posed for promotional images, performed a small number of demonstration tasks involving simple object manipulation, and was returned to Earth aboard the same Soyuz capsule after approximately two weeks. The Russian space program has not announced a follow-on mission. The program has, in operational terms, gone dark since 2019, with Roscosmos’s broader budget pressures and the post-2022 Western sanctions regime making any near-term follow-on extremely unlikely.

The Chinese side of the humanoid-in-space history is, as of public disclosure, minimal. China Manned Space Engineering Office (CMSEO), which operates the Tiangong space station, has not publicly launched a humanoid robot to Tiangong. The station’s robotic capabilities are concentrated in a Tiangong robotic arm system modeled architecturally on the Canadarm design used on the ISS, with associated smaller manipulator arms for crew-internal use. Various Chinese commercial humanoid manufacturers — Unitree, AgiBot, Fourier Intelligence, UBTECH — have discussed long-term space-deployment ambitions, but no Chinese humanoid robot has flown a space mission as of public reporting in 2026.

What’s actually working in space: Astrobee, Int-Ball, and the free-flying drone category

The robotics platforms doing real operational work on the ISS in 2026 are not humanoid. They are free-flying cubes. Astrobee, a NASA Ames Research Center program that delivered three robots — Honey, Queen, and Bumble — to the ISS in 2019, are cube-shaped autonomous flying robots approximately 12.5 inches on a side, propelled by electric impeller fans that move air to generate thrust in microgravity, and equipped with cameras, displays, and a robotic perching arm that allows the robot to attach to handrails or other ISS interior fixtures for stable observation. Astrobee operates as a free-flying assistant performing routine surveys, inventory tracking, environmental monitoring, and as a mobile platform for hosting visiting research payloads from external university and commercial users. The platform has accumulated thousands of hours of autonomous operation on the ISS since 2019, more than any humanoid robot has ever accumulated in space.

The Japanese counterpart is Int-Ball, a spherical free-flying camera drone developed by JAXA’s Japan Aerospace Exploration Agency and deployed to the ISS Japanese Experiment Module (Kibo) in 2017, with a successor Int-Ball 2 delivered to the station in 2024 with improved autonomous-navigation capability and higher-resolution video. The German-Airbus-IBM collaborative platform CIMON (Crew Interactive Mobile Companion), a softball-sized AI-powered free-flying assistant equipped with conversational interface software, has flown two ISS missions since 2018 with European astronaut Alexander Gerst and subsequent crew. The structural commonality across Astrobee, Int-Ball, and CIMON is that none of them have legs, none of them are anthropomorphic, none of them attempt to mimic the human form factor, and all of them have substantially more operational hours in space than the entire global humanoid-robot fleet combined.

The Mars rotorcraft revolution: Ingenuity and its successors

The most consequential aerial robotics platform ever deployed beyond Earth’s atmosphere is Ingenuity, the NASA JPL twin-rotor Mars helicopter that flew alongside the Perseverance rover after the rover’s February 2021 landing in Jezero Crater. Ingenuity was, by the program’s original design parameters, a technology demonstration intended to prove the feasibility of powered atmospheric flight on Mars across a five-flight, thirty-day primary mission. Ingenuity flew its first powered, controlled flight on Mars on April 19, 2021 — the first time a vehicle had performed powered flight on another planet — and then proceeded to massively exceed its design specification. The helicopter accumulated 72 successful flights over 33 months of operations, flew a cumulative total of approximately 17 kilometers across the Martian surface, reached maximum altitudes of approximately 24 meters above the ground, and performed scouting flights for the Perseverance rover that materially affected the rover’s traverse planning. Ingenuity’s final flight occurred on January 18, 2024, at a location JPL informally designated Valinor Hills, when the helicopter sustained rotor-blade damage on landing that ended its ability to fly. The mission was concluded shortly thereafter.

The Mars helicopter program is being expanded under the Mars Sample Return mission architecture, with two Sample Recovery Helicopters planned for the mid-to-late 2020s as backup retrieval vehicles for the Perseverance sample cache. The Dragonfly mission, scheduled for launch in 2028 and arrival at Saturn’s moon Titan in 2034, is an eight-rotor electric drone built by the Johns Hopkins University Applied Physics Laboratory that will fly across Titan’s nitrogen-methane atmosphere — denser than Earth’s at a tenth the gravity — to perform geological and astrobiological surveys at multiple landing sites. The rotorcraft category, in 2026, is the most successful new-form-factor robotics platform ever introduced into the planetary exploration architecture. Every Ingenuity flight on Mars produced more rigorous public-attention data on autonomous robotics than every NASA humanoid program combined.

The commercial lunar lander wave: Blue Ghost, Athena, Peregrine, and the partial-success era

The commercial lunar lander category — operating under NASA’s Commercial Lunar Payload Services (CLPS) program, which awards relatively cheap fixed-price contracts to private-sector companies to deliver scientific payloads to the lunar surface — has, since January 2024, produced a sequence of partial-success and failure outcomes that have characterized the practical state of robotic lunar landing in 2026. Astrobotic‘s Peregrine Mission One launched in January 2024 and failed in transit due to a propellant leak; the spacecraft was deliberately re-entered into Earth’s atmosphere without reaching the Moon. Intuitive Machines‘s IM-1 Odysseus launched in February 2024 and made the first commercial soft landing on the lunar surface, but the spacecraft tipped over on touchdown and ended its mission early. Firefly Aerospace‘s Blue Ghost Mission 1 (“Ghost Riders In the Sky”) launched in January 2025 and on March 2, 2025 completed the first fully-successful vertical landing of a U.S. spacecraft on the lunar surface since Apollo 17 in December 1972, with Will Coogan serving as Firefly’s chief engineer for the lander. Blue Ghost operated near the lunar equator in Mare Crisium with ten NASA-sponsored instruments, including the Lunar PlanetVac sample-acquisition system built by Honeybee Robotics. Intuitive Machines‘s IM-2 Athena launched in February 2025 and landed on March 6, 2025 near the lunar south pole, approximately 820 feet (250 meters) from its intended landing site, with the spacecraft again ending in a non-nominal attitude that ended the mission prematurely. NASA paid Firefly approximately $101 million for the Blue Ghost delivery contract, with an additional $44 million for the instruments themselves. Astrobotic’s Griffin Mission One is planned for no earlier than December 2025, with Blue Ghost Mission 2 and IM-3 scheduled for subsequent windows. The lunar lander category is, in 2026, the most active commercial space-robotics market in the world, with multiple U.S. private-sector companies competing aggressively for NASA CLPS contracts.

The GITAI lunar-rover and ISS robotic-arm story

The lunar surface robotics category in 2026 is dominated by a Japanese-American space robotics company called GITAI, founded in Tokyo and now headquartered in Torrance, California with a Japanese subsidiary called GITAI Japan Inc. The company’s signature platform is the Inchworm robotic arm — a modular, segmented manipulator designed to “walk” along structural attachment points by alternately attaching and detaching at its two endpoints, allowing the arm to relocate itself across a spacecraft’s exterior or a lunar surface installation without requiring a separate locomotion system. GITAI has completed successful technical demonstrations of robotic arms both inside and outside the ISS, including a 1.5-meter dual-arm S2 system that completed an external ISS demonstration of autonomous structure-assembly and maintenance tasks. In June 2024, GITAI was selected for NASA’s Small Business Innovation Research (SBIR) Phase 1 program. In January 2025, GITAI completed a space demonstration of its in-house developed 16U-class CubeSat in low Earth orbit, validating attitude control and propulsion systems. In March 2025, JAXA awarded GITAI Japan a concept-study contract for the robotic arm system on Japan’s contribution to NASA’s Artemis program — the pressurized crewed lunar rover that will support long-duration human exploration of the lunar south polar region. In April 2025, GITAI established a U.S. defense-and-space subsidiary called GITAI Defense and Space LLC to expand U.S. government contracting capabilities. The Inchworm arm has, as of public disclosure, completed environmental testing including regolith exposure, thermal vacuum, vibration, and radiation tests sufficient to achieve Technology Readiness Level 6 (TRL-6) for operations in the lunar south polar environment.

The orbital servicing and debris-removal category

The orbital servicing category — robots that approach existing satellites in geostationary or low-Earth orbit and perform refueling, repair, or controlled deorbiting operations — is dominated by two companies in 2026. Northrop Grumman SpaceLogistics operates the Mission Extension Vehicle (MEV) platform, with MEV-1 docked to the defunct Intelsat-901 geostationary satellite in February 2020 and providing operational life extension for five additional years, and MEV-2 docked to Intelsat 10-02 in April 2021. The MEV platform is, as of 2026, the most successful commercial orbital-servicing platform ever deployed. Astroscale, a Japanese-British orbital-servicing company, operates the ADRAS-J spacecraft, which in 2024 conducted the world’s first detailed close-proximity inspection of a defunct rocket stage — a Japanese H-IIA upper stage that had been in orbit since 2009 — and demonstrated the rendezvous and proximity-operations capability needed for active debris removal. ClearSpace SA, a Swiss orbital-servicing company contracted by the European Space Agency, is developing ClearSpace-1, a debris-removal spacecraft intended to capture and deorbit the VESPA upper stage. Orbit Fab is developing the GAS Station for Satellites orbital-fuel-depot infrastructure to enable refueling-based satellite life extension. The orbital servicing market is, in 2026, in the same approximate stage of commercialization that maritime autonomy was in 2020 — a small number of operational platforms, a growing set of demonstrated capabilities, and a market that institutional customers (commercial satellite operators, defense agencies, space-debris-conscious regulators) are slowly beginning to take seriously.

The Mars and lunar surface rovers

The wheeled-rover category, the most operationally mature robotic platform in deep-space exploration, continues to operate in 2026 with multiple platforms across multiple destinations. NASA’s Perseverance rover, which landed in Jezero Crater on Mars in February 2021, continues to traverse Jezero’s western delta with the Ingenuity companion now retired at Valinor Hills. The Curiosity rover, operating in Gale Crater since August 2012, continues to climb Mount Sharp with continued instrument operation more than a decade past its original two-year primary mission. The China National Space Administration‘s Zhurong rover, which landed on Mars in May 2021 as China’s first interplanetary lander, completed its primary mission and entered hibernation in May 2022; the rover has not transmitted since and is presumed to have failed during a Martian winter. The India Space Research Organisation‘s Pragyan rover, deployed by the Chandrayaan-3 lunar lander in August 2023 in the lunar south polar region, operated for one lunar day before lunar night ended its mission. China’s Chang’e-6 mission returned the first samples from the lunar far side in June 2024. The rover category is, in 2026, what the agricultural and mining robotics markets were in 2010 — a mature, operationally-proven, mission-essential platform category that has long since left the technology-demonstration phase.

Power, payload, and the brutal physics of off-Earth operation

The physics constraints that define what can and cannot operate as a robot in space are unforgiving. Every kilogram launched to low Earth orbit costs between $1,500 (SpaceX Falcon 9) and $10,000 (legacy expendable launch vehicles). Every kilogram launched to lunar orbit costs roughly five to ten times the LEO figure. Every kilogram landed on the Martian surface costs roughly twenty to fifty times the LEO figure. A robot designed for terrestrial deployment can carry a 50-kilowatt-hour battery and recharge daily. A robot designed for Martian deployment must operate on solar arrays delivering, at best, a few hundred watts during daylight hours, or must carry a radioisotope thermoelectric generator (RTG) powered by plutonium-238 — the same isotope category that powers Curiosity and Perseverance — at extreme cost and with extreme supply-chain constraints (the U.S. plutonium-238 production capacity is, in 2026, less than 1.5 kilograms per year, against a Mars-rover requirement of approximately 4.8 kilograms per rover). Radiation outside Earth’s magnetosphere degrades semiconductor electronics on timescales of months to years rather than the decades that terrestrial chips routinely operate for, requiring radiation-hardened components that cost orders of magnitude more than commercial equivalents and that lag commercial computing performance by approximately a decade. Vacuum demands seals, lubricants, and thermal management systems that have no terrestrial analog. Microgravity changes the physics of every fluid system in the spacecraft.

These constraints explain why space robotics has not converged on humanoid platforms. The robot that makes operational sense in space is the one optimized for the specific environmental constraints of the specific destination — a free-flying cube for the ISS interior, a rotorcraft for thin atmospheres, a wheeled rover for planetary surfaces, an inchworm arm for orbital structures, a dedicated docking-and-grappling spacecraft for orbital servicing. The space robotics industry has, over six decades, learned that the human body is not the natural form factor for off-Earth operation. The recent humanoid-robot enthusiasm that has driven the commercial humanoid-robot race on Earth does not extend, in any meaningful operational sense, to actual space deployment.

What 2026 looks like across space humanoid robots and drones

In 2026, the operational reality of robots in space is dominated by non-humanoid platforms doing non-humanoid work. The ISS interior is patrolled by Astrobee cubes, Int-Ball cameras, and CIMON conversational assistants. The ISS exterior is serviced by the Canadarm2 robotic arm (Canadian Space Agency, operational since 2001) and the smaller Dextre manipulator. The Martian surface is operated by Perseverance and Curiosity rovers, with Ingenuity retired and successor rotorcraft in development, alongside ongoing concept studies for quadrupedal lunar and Martian surface platforms based on Spot-derived hardware under JPL and DLR research programs. The lunar surface is contested by U.S. commercial landers (Firefly Blue Ghost successful, Intuitive Machines IM-1 and IM-2 tipped, Astrobotic Peregrine failed in transit) under NASA’s CLPS program, with Astrobotic Griffin, Blue Ghost Mission 2, IM-3, and Japan’s ispace Resilience missions in pipeline. Orbital servicing is operated by Northrop Grumman MEV-1 and MEV-2, with Astroscale ADRAS-J demonstrating debris-inspection capability, and with ClearSpace-1 and Orbit Fab infrastructure in development. Humanoid robots, despite the Apptronik Apollo NASA-Valkyrie genealogy and the Tesla Optimus Mars-deployment promises that Elon Musk has periodically made in public communications, have functionally zero deployed operational presence in space in 2026. The Russian Skybot FEDOR program has gone dark. NASA’s Robonaut 2 sits on the ground at Johnson Space Center. NASA’s Valkyrie remains a ground-based research platform. The Chinese Tiangong station has not received a humanoid robot.

The gap between the commercial humanoid-robot industry’s NASA-leveraged marketing and the actual humanoid presence in space tells a useful story about how the space robotics industry has evolved over six decades. The constraint set — cost-per-kilogram, radiation environment, microgravity, vacuum, lack of maintenance windows — has driven the platform architecture in directions that have nothing to do with the human body. The robots doing the most operationally consequential work beyond Earth are cubes, rotorcraft, rovers, arms, and dedicated servicing spacecraft. The robots that make for the most compelling marketing imagery — bipedal humanoids standing on the lunar surface, working alongside astronauts on a Mars base, performing maintenance on a space station — are, with the exception of the brief and limited Robonaut 2 and Skybot FEDOR demonstrations, theoretical. The humanoid-robot industry on Earth continues to expand at the pace its commercial customers and venture investors are willing to fund. The space humanoid-robot industry is, in operational terms, a category that has not yet meaningfully begun.

Whether that changes depends on three structural variables. The first is the long-term cost trajectory of launch — if SpaceX Starship achieves its public design target of $100/kg to low Earth orbit, the economics of launching heavy humanoid platforms shifts by an order of magnitude and the marginal cost of putting an Apollo unit on the Moon becomes plausible rather than prohibitive. The second is the long-term trajectory of human spaceflight — if NASA’s Artemis program and the various commercial space-station ventures (Axiom Space, Vast Space, Voyager Space’s Starlab) actually scale into operational platforms with consistent crew presence, the operational case for humanoid robots that share environmental design with the human crew becomes stronger. The third is the long-term trajectory of the humanoid-robot industry itself — if Apptronik, Tesla, Figure, Agility Robotics, Boston Dynamics, and the broader commercial humanoid cohort actually scale their platforms into reliable, low-maintenance, factory-floor-grade industrial robots, the marginal engineering effort required to space-qualify a unit becomes meaningful rather than speculative.

None of those three structural variables is on its own trajectory to resolve in the near term. Starship has not yet achieved orbital flight at the cost and reliability targets the company has publicly committed to. Artemis has not yet landed a crewed mission on the lunar surface, with Artemis II scheduled for crewed lunar flyby in 2026 and Artemis III scheduled for the first crewed landing in 2027. The commercial humanoid robot industry continues to scale aggressively but has not yet demonstrated the operational reliability — the Tuesday-proof, not-babysat-by-PhDs deployment — that would justify the additional engineering investment to space-qualify a platform. The humanoid-in-space narrative is, in 2026, a marketing story leveraging genuine NASA pedigree to sell terrestrial products. The actual robots doing work beyond Earth are cubes, rotorcraft, rovers, arms, and servicing spacecraft, operating under the same supply chain constraints, the same software-development practices, and the same evolving regulatory architecture that govern the broader commercial robotics industry — but operating in physical environments that have, six decades into the space age, definitively converged on form factors that have nothing to do with the human body. The robots in space, like the robots in companionship applications, are the result of long convergence between what the technology can do and what the environment will tolerate. Six decades of space operations has produced an answer that is, in 2026, more confident about what doesn’t work than about what eventually will.
Forestry, Land Management and Conservation Robotics in 2026: The Hardest Robotics ROI to Verify

On a moonless night in late 2014, a small fixed-wing drone equipped with an infrared thermal imager lifted off from a ranger station in the Pretoriuskop section of Kruger National Park in northeastern South Africa, climbed to its operating altitude of roughly 100 meters, and began flying a programmed search pattern across roughly fifty square kilometers of scrub bush, dry riverbeds, and sparse miombo woodland. The drone’s pilot — a former park ranger named Graham Dyer, operating under a six-week trial contract — sat in front of a laptop in the ranger station, watching the thermal feed for the distinctive double signature that indicates a human figure on foot near a rhinoceros. The rhinoceros warms the savanna with the radiative pattern of a 3,000-pound mammal. The human shows up as a smaller, sharper, often-moving heat source against the same background, typically carrying a rifle. The drone records both signatures, transmits the coordinates back to the ranger station, and the on-foot patrol team is dispatched to interdict. Over the six weeks of the Pretoriuskop trial, while the drone was airborne, no rhinos were killed. In the previous month, in the same area, without the drone, nine rhinos had been poached.

This is the domain where the robotics industry’s environmental and conservation claims are stress-tested against the hardest possible measurement environment. Kruger National Park covers 19,485 square kilometers — roughly the size of Wales — and at the peak of the South African rhino-poaching crisis between 2013 and 2015, approximately 1,400 rhinos were being killed per year, an average of three to four per day. By 2020, that rate had fallen to one rhino killed approximately every 22 hours. By 2024, it had declined further, with a combination of armed patrols, dehorning programs, thermal-equipped drones, AI-based monitoring, and rhino relocation jointly responsible for the recovery. The conservation-drone fleet — Air Shepherd, a Lindbergh Foundation program that has flown over 4,000 missions across South Africa, Malawi, and Zimbabwe; the Hluhluwe/iMfolozi Park anti-poaching unit’s AI-and-thermal systems in KwaZulu-Natal; and a long tail of smaller park-specific deployments — is the most credible operational success story in the conservation-robotics category. The technology originally developed for U.S. military roadside-bomb detection in Iraq has been repurposed, with the same hardware family and the same image-processing algorithms, to do the exact opposite of what the autonomous-weapons industry is building it for — to detect humans who are about to kill animals, rather than to kill humans before they detect the drone.

The reforestation drone wave and the Mast pivot

In late 2016, a Seattle-based startup called DroneSeed — founded by Grant Canary, the CEO who had previously cycled through Techstars Seattle’s 2016 cohort — launched the most publicized application of robotics to climate-change mitigation that the industry had attempted: drone-swarm aerial reseeding of forested land destroyed by wildfire. The model was elegant on paper. The United States loses an average of 70,000 wildfires and 7 million acres of forest per year. Natural regeneration is slowing as wildfires get hotter and more frequent. Hand-planting reforestation crews are constrained by manual-labor scaling limits and a 2-to-3-year seedling supply chain bottleneck. A swarm of heavy-lift drones, each carrying a 57-pound payload of engineered “seed pucks” containing pine seeds, fertilizer, and a moisture-retention substrate, could in principle drop the supply chain bottleneck from 3 years to 3 months, plant tens of thousands of acres in days rather than seasons, and finance the whole operation through carbon credits sold to corporate buyers under the voluntary carbon market.

DroneSeed was the only reforestation company FAA-approved to fly drones with payloads above 55 pounds, to fly drones in swarms, and to fly drones beyond visual line of sight — a regulatory advantage that, in the parallel agricultural-drone market, would have been worth a significant valuation premium. The company rebranded as Mast Reforestation in 2023 (named for the forestry term mast years, the infrequent years when trees produce bumper crops of seed cones), acquired Silvaseed — a 130-year-old Western Washington seed bank that was the largest private seed supplier west of Colorado — in 2021, acquired Cal Forest Nurseries to become the largest seed-and-seedling vendor in the western United States, and built out a vertically-integrated pipeline that paired drone-deployed seed pucks with traditional hand-planted seedlings. By 2023, Mast had replanted approximately 2,500 acres of Montana and had a project pipeline of 20,000 additional acres. In February 2025, Mast closed a $25 million Series B round co-led by Chamath Palihapitiya‘s Social Capital, bringing total funding to roughly $81.74 million.

The operational results have, as of 2026, been substantially worse than the model predicted. In January 2025, Mast informed its partner Carbon Streaming that the drone- and hand-planted seedlings at the Sheep Creek, Baccala Ranch, and Feather River Dome projects had “experienced significantly higher than expected mortality rates and that the surviving seedlings had exhibited slower than expected growth rates.” Mast quietly withdrew several rounds of “forecasted mitigation units” — pre-sold carbon credits priced against the projected sequestration of planted seedlings — from the voluntary carbon market when the underlying biology failed to materialize. By June 2025, Mast was facing a fraud lawsuit from a former project partner. By February 2025, the company had pivoted its core business model from drone-and-hand reforestation to biomass burial — burying dead, fire-killed trees in clay-rich pits to prevent decomposition and trap their carbon underground — and announced the pivot alongside the Series B fundraise. The most ambitious conservation-robotics company of the 2016-2024 era is, in 2026, a tree-burial company that still does some drone seeding on the side. The promise the drones encoded — that you could mechanize reforestation at scale and finance it through carbon markets — has, structurally, not survived contact with the seedlings.

The post-Mast reforestation-drone ecosystem continues. Flash Forest in Canada operates a similar drone-seed-pod model focused on boreal reforestation. Dendra Systems (formerly BioCarbon Engineering), founded by ex-NASA engineer Lauren Fletcher, operates ecosystem-restoration drone projects in the United Arab Emirates, Australia, the United Kingdom, and Madagascar. AirSeed Technologies in Australia operates a drone-seed model focused on Australian native species and post-bushfire restoration. The combined deployed footprint is, by 2026, somewhere in the hundreds of thousands of acres treated cumulatively — a small fraction of the 70 million acres burned in the United States alone over the last decade, and a smaller fraction still of the global reforestation need. The technology works at the level of individual seed dispersal. The financial model that would scale it to the size of the problem has not yet emerged.

The anti-poaching drone and the night-vision arms race

The anti-poaching domain, by contrast, has been the conservation-robotics category with the cleanest operational evidence. Air Shepherd — formally part of the Charles A. and Anne Morrow Lindbergh Foundation — uses fixed-wing drones equipped with thermal-imaging cameras, originally developed for the U.S. military’s Iraq-era roadside-bomb-detection program, to fly nighttime patrols across high-poaching-probability zones in South African, Malawian, and Zimbabwean national parks. The drones operate primarily at night because approximately 80% of all poaching occurs in the hours of darkness. The thermal-imaging systems can distinguish the heat signature of a human carrying a rifle from the surrounding bush and animal heat. The on-the-ground response is conducted by armed park rangers; the drone is the detection layer, not the interdiction layer. As of 2026, Air Shepherd has operated over 4,000 patrol missions.

The operational impact, while difficult to attribute cleanly because the anti-poaching campaign has involved many parallel interventions (rhino dehorning, intelligence-led arrests, increased patrol funding, K-9 units, demand-reduction campaigns in Vietnam and China), is at minimum strongly correlated with a sustained decline in South African rhino mortality. The peak of approximately 1,400 rhinos poached per year in 2014 had declined by roughly 60-70% by 2024. Crawford Allan, the World Wildlife Fund’s crime-technology project spokesperson, has publicly described Kruger as “ground zero for poachers,” with as many as twelve organized poaching gangs operating inside the park at any given time. The conservation-drone fleet has, in the operational reading of the WWF and the South African National Parks (SANParks) leadership, contributed materially to the reduction. The same family of camera-and-autonomy technology that runs the DFR drone programs at Chula Vista PD is, in Kruger, watching rhinoceroses sleep — a structural reuse of the same Skydio and DJI-derived platform stack that has scaled into every other drone-deployment domain in the cluster. The hardware stack depends on the same semiconductor supply chain, the same lithium-ion battery chemistry, and the same rare-earth permanent magnets in the motors as every other autonomous platform the cluster has documented — including the same Boston Dynamics Spot platforms that several South African private game reserves have, since 2024, begun acquiring for perimeter patrol and night-time inspection of remote ranger outposts.

The conservation-drone story extends well beyond anti-poaching. South African conservationist Carel Verhoef in 2024 used a small fleet of drones and ranger pilots to move a herd of 150 elephants 70 kilometers at night across the Tanzania-Kenya border, using the drones as a noise-and-presence shepherding tool to redirect the herd away from a corridor where they were vulnerable to poaching and toward a protected reserve. Chisl/Veriphy AI, a Johannesburg-based group founded by Willem Kellermann, conducted a major drone-based wildlife census in 2025 covering more than 100,000 hectares in several private game reserves near Kruger, using AI-driven image processing to count elephants, rhinos, buffalo, antelope, and lions at a fraction of the cost of historical helicopter-based aerial census methods. Ezemvelo KZN Wildlife in KwaZulu-Natal flies BVLOS drones for both rhino-monitoring and rare-plant work — including a multi-year project to locate the so-called “loneliest plant in the world,” a single specimen of Encephalartos woodii believed to be the last of its species, using a combination of drones, satellites, and spectral imaging. The conservation-drone footprint across sub-Saharan Africa is, by 2026, somewhere in the low thousands of operational airframes across hundreds of parks and reserves.

RangerBot and the Great Barrier Reef

In August 2018, after winning the $750,000 People’s Choice prize at the 2016 Google Impact Challenge, researchers from Queensland University of Technology under principal investigator Matthew Dunbabin launched RangerBot at the Reef HQ Aquarium in Townsville, Queensland. RangerBot is a 15-kilogram autonomous underwater vehicle with six thrusters, two stereo camera systems for visual navigation, and a single dedicated function: identify and inject the crown-of-thorns starfish (COTS), the invasive coral-eating echinoderm whose population booms across the Great Barrier Reef have, since the early 2010s, been one of the most consequential drivers of coral loss after thermal bleaching. RangerBot identifies COTS with 99.4% accuracy using onboard computer vision, dispatches a lethal dose of vinegar or bile salts via injection arm into each identified specimen, and operates for eight hours on a single charge — roughly three times longer than a human diver can stay below the surface.

The structural argument for RangerBot was scale economics. The Great Barrier Reef Marine Park Authority (GBRMPA) reported that across 2023-2024, 16,657 hours of human-diver effort culled approximately 50,227 COTS — a rate of one starfish killed every 20 minutes. A fleet of RangerBots, each operating continuously for eight-hour shifts and identifying COTS in real time, could in principle achieve culling rates an order of magnitude higher than the diver-based baseline. The actual operational deployment, as of 2026, remains in the low-single-digit-fleet-size range — RangerBot is built in QUT laboratories rather than mass-produced by a commercial manufacturer, and the GBRMPA’s COTS-control program remains predominantly diver-based. The complementary Down Deep Drones prototype, built by an independent Australian developer for approximately $6,000 on an off-the-shelf QYSEA underwater drone platform, has been pitched to GBRMPA and the Reef and Rainforest Foundation since 2018 with mixed reception. The technology works on a per-starfish basis. The institutional adoption pathway that would scale it to the size of the COTS outbreak has not closed.

The broader underwater-conservation-robotics ecosystem includes LarvalBot (a sister project at QUT that dispenses coral larvae onto degraded reefs to accelerate regeneration), Mesobot at the Monterey Bay Aquarium Research Institute (which tracks individual zooplankton at midwater depths for ocean-research purposes), and a growing fleet of academic-research AUVs operating in the same family of low-cost commercial platforms — OpenROV Trident units, QYSEA FIFISH professional models, and the Blue Robotics BlueROV2 — that have made underwater robotics accessible to research budgets that could not previously afford an oceanographic-grade ROV. The combined deployed footprint of conservation-and-research AUVs across global coastal-management programs is, by 2026, in the low tens of thousands of units, dominated by the consumer-grade Chinese platforms and the academic-grade U.S. and European systems.

Forest inventory, LiDAR drones, and the timber supply chain

The commercially largest application of drones in the broader land-management category is forest inventory — the cataloging of standing timber, biomass density, species mix, and harvestable volume across managed and unmanaged forests for the timber, paper, carbon-credit verification, and forest-management industries. Treeswift, a Philadelphia-based startup, operates a fleet of LiDAR-equipped autonomous drones that fly under forest canopy to inventory individual trees, identify species, and measure trunk diameter at breast height — work that historically required ground crews with handheld measuring tape and clipboards. Sweden’s Skogforsk research institute operates a comparable program for the Scandinavian timber industry. Finland’s Metsähallitus flies drones for state-forest inventory. The U.S. Forest Service operates several thousand drones across its 193-million-acre management portfolio for fire-line monitoring, post-fire assessment, invasive-species surveys, and recreation-area management. The Bureau of Land Management operates a parallel fleet across the 245 million acres under its jurisdiction.

The forest-fire-monitoring side of the land-management domain bleeds directly into the autonomous wildfire-suppression aircraft documented in the firefighting cluster post — the Sikorsky-Rain autonomous Black Hawk that conducted live-fire suppression tests in April 2025 is, in operational terms, the upper end of the same fire-monitoring-and-suppression continuum that smaller drone fleets occupy at the lower end. Pano AI, a San Francisco startup that operates a network of high-mountain cameras for early wildfire detection, integrates with state-fire-agency drone-dispatch systems across California, Oregon, Washington, Colorado, and several other Western states. The combined real-time wildfire-monitoring fleet across the U.S. West — drones, fixed cameras, satellite-based hot-spot detection, and crewed reconnaissance aircraft — has dramatically reduced the average time between fire ignition and first response over the last decade, with corresponding measurable reductions in average burn area for fires detected in the first hour.

The wildlife census and the disappearing penguin

The most consistently funded and operationally successful category of conservation drone is the wildlife population census. The British Antarctic Survey has, since 2017, used fixed-wing drones to count penguin colonies across the Antarctic Peninsula, South Georgia, South Orkney, and the South Sandwich Islands — work that historically required ship-based expeditions counting from binoculars and which the drones now accomplish in fractions of the time at a fraction of the cost. The University of Sydney‘s wildlife-monitoring drone program counts kangaroo, wallaby, and koala populations across New South Wales and Queensland. The U.S. National Park Service flies drones for Yellowstone bison counts, Glacier bighorn sheep counts, and Channel Islands fox monitoring. The University of Cape Town‘s African Penguin Initiative uses drones to count breeding colonies along the South African coast — a population that has, despite the monitoring, declined by more than 60% since 2000 and is now classified as critically endangered.

The structural distinction in the wildlife-census category is that the drone is a measurement instrument rather than an intervention. The robot does not change the population. It tells the conservation manager what the population is. The decisions about whether to relocate animals, install electric fencing, deploy anti-poaching patrols, or close fisheries to protect prey species are downstream of the data. The conservation outcome depends on the institutional capacity to act on the measurement. This is the recurring constraint in every conservation-robotics deployment the cluster has documented — the robots can do the surveillance and the intervention, but the conservation outcome depends on the political, legal, and financial framework around them. The Air Shepherd drone identifies the poacher. The on-foot ranger team has to make the arrest. The RangerBot identifies the COTS. The GBRMPA management plan has to scale the deployment. The Mast Reforestation drone drops the seed puck. The seedling has to survive the next three drought summers.

Marine conservation and the Saildrone fisheries program

The largest operational deployment of autonomous vehicles in marine conservation in 2026 is the NOAA Fisheries program that uses Saildrone Voyager units for trawl-survey calibration, salmon-population assessments off the U.S. West Coast, pollock-population assessments in the Bering Sea, and acoustic monitoring of cetacean populations across the U.S. EEZ. Saildrone has, as of 2026, completed multi-year contracts with NOAA, with the U.S. Coast Guard for civilian and dual-use missions, and with the Australian Bureau of Meteorology for Pacific climate monitoring. The vessels are the same 23-foot solar-and-wind-powered platforms that the maritime defense industry has scaled for U.S. Navy task force operations — the dual-use overlap is total. The same Voyager that maps a Bering Sea pollock biomass survey in March can be re-tasked for Replicator maritime-domain-awareness in the South China Sea in June with no hardware modification.

The fisheries-assessment use case is, in conservation-robotics terms, the strongest published-evidence example outside of African anti-poaching. NOAA’s Saildrone-based pollock surveys have, in head-to-head comparison studies against traditional crewed fishing-vessel-and-acoustic-transducer assessments, produced comparable biomass estimates at substantially lower cost and with substantially less impact on the surveyed fish populations. The structural argument for the autonomous platform is the same as it is in every other robotic-deployment domain in the cluster: the unit cost is lower, the duration is longer, the human risk is lower, and the data quality is, in some categories, measurably better.

What 2026 looks like in conservation robotics

In 2026, Air Shepherd’s anti-poaching drones continue to fly across South Africa, Malawi, and Zimbabwe, with the broader anti-poaching technology ecosystem — thermal imaging, AI-driven image processing, BVLOS regulatory waivers, integrated ranger dispatch — credited with material contribution to the ~60-70% decline in South African rhino mortality since the 2014 peak. Mast Reforestation continues to operate as a hybrid drone-seeding-and-biomass-burial business, with the original drone-swarm reforestation model having largely failed against its carbon-credit projections, and a fraud lawsuit pending against the company. Flash Forest, Dendra Systems, and AirSeed Technologies continue to operate smaller reforestation-drone programs in Canada, the UAE/Australia/UK/Madagascar, and Australia, respectively. RangerBot continues to be deployed in limited fleet sizes at the Great Barrier Reef alongside the larger diver-based COTS-control program. Treeswift, the U.S. Forest Service, the Bureau of Land Management, and a constellation of state and private timber-industry operators run a forest-inventory drone fleet measured in the low tens of thousands of airframes. NOAA’s Saildrone fisheries-assessment program continues to expand. The British Antarctic Survey, the U.S. National Park Service, and a long tail of academic wildlife-census programs continue to operate drone-based population counts that have replaced helicopter-and-binocular-based methods at orders-of-magnitude lower cost.

The conservation-robotics category does something the rest of the cluster has not asked the technology to do — it asks the robot to be the substitute for institutional capacity that the conservation movement has not been able to build at scale. The reforestation drone was supposed to replace the manual planting crew that the forestry industry cannot afford to scale. The anti-poaching drone was supposed to replace the ranger patrol that the African national parks cannot fund to the size of their territories. The RangerBot was supposed to replace the human diver who cannot stay submerged long enough to keep up with the COTS outbreak. The wildlife-census drone was supposed to replace the helicopter survey that no national park system in the world has budgeted at the frequency the science requires. In each case, the robot does the work the human alternative cannot do — and in each case, the binding constraint on the conservation outcome is not the robot’s capability but the institutional structure around it. The carbon-credit market has not been able to verify the Mast Reforestation projects’ biological outcomes. The South African rhino population is recovering not because the drone alone interdicts the poacher, but because the drone’s detection feeds an armed ranger team that the South African government has been willing to staff and arm at scale. The Great Barrier Reef’s COTS population is not falling fast enough because the RangerBot fleet is not big enough, because the GBRMPA budget is not large enough, because Australian climate policy has not, in the operational reading of the marine-biology community, addressed the underlying nutrient-runoff and thermal-bleaching pressures that drive the COTS outbreak in the first place.

The robots in this cluster are, in some ways, the cluster’s most morally compelling deployments — the Spot patrolling an offshore oil platform is not saving an endangered species, the Trajekt Arc throwing 100-mph cutters in a basement batting cage is not buying time for a coral reef, and the humanoid robot demoing on a stage at a venture-capital conference is not, in any direct sense, addressing the biosphere collapse that the conservation-robotics community has spent the last fifteen years building hardware against. The conservation drone, the anti-poaching thermal imager, the reforestation seed puck, the underwater starfish-injector, and the autonomous fisheries-assessment platform are the rare robots whose mission statement is, structurally, “do something the planet’s biosphere desperately needs.” The fact that the conservation-robotics category has the most ambitious mission and the most mixed operational evidence is not, in the cluster’s running thesis, a failure of the robots. It is a failure of the institutional framework around the robots — the carbon markets, the national park budgets, the international wildlife-trade enforcement regimes, the climate-policy frameworks, the conservation-infrastructure budgets that no national government has been willing to fund at scale — to match the capability of the underlying robotic platforms the scientific research community and the K-12-to-university talent pipeline have spent decades producing — including the deliberately-cute consumer-facing robots whose design budgets, in 2026, dwarf the entire global conservation-robotics R&D spend by a factor of perhaps fifty to one. The robots will keep doing the work. Whether the planet’s ecosystems recover enough to justify having built them is, in 2026, still being decided by the institutions the robots cannot, by themselves, fix — and the gap between the robotic capability and the conservation outcome remains, across every domain the cluster has documented, the most morally consequential and the least technologically solvable problem in the entire industry.
Scientific Research and University Robotics in 2026: Where the Hard Robots Get Invented

On January 18, 2024, at approximately 12:00 UTC, a four-pound tissue-box-sized helicopter named Ingenuity lifted off from the dust of Jezero Crater on Mars for the seventy-second time, climbed to twelve meters of altitude, hovered briefly, and descended for what its operators at NASA’s Jet Propulsion Laboratory expected to be a routine systems-check landing. Somewhere in the final meters of descent, the helicopter’s downward-facing navigation camera lost track of the featureless sand-rippled terrain below it, the autonomous flight controller misjudged the height and ground speed, and Ingenuity touched down hard enough to damage at least one rotor blade — a “blade strike,” in the language of rotorcraft engineering, that on a Martian helicopter with no spare parts and no maintenance crew is functionally equivalent to a terminal diagnosis. JPL’s project manager Teddy Tzanetos confirmed the helicopter would fly no more. The team named the spot Valinor Hills, after the fictional location in J.R.R. Tolkien’s legendarium where the Elves go to die. Ingenuity had been designed for a five-flight, thirty-day technology demonstration. It flew 72 missions across nearly three years, covered roughly 17 kilometers in total, and proved for the first time in human history that powered, controlled, atmospheric flight was possible on another planet. It still transmits weather data to the Perseverance rover roughly once per week. The most expensive single autonomous helicopter ever built — at approximately $85 million across its design, fabrication, integration, and operations through the demonstration phase — is now a memorial in the floor of an impact crater on Mars, marking the upper boundary of what the scientific research robotics community can build when the timeline is two decades, the budget is a NASA appropriation, and the objective is to demonstrate that something previously thought impossible is in fact possible.

This is the domain where the hard robots get invented. The humanoid robots that Figure AI and Apptronik are deploying into commercial pilot programs, the autonomous helicopters that Sikorsky and Rain are testing against California wildfires, the agricultural sprayers that Carbon Robotics and Hylio are flying across Iowa cornfields, the autonomous container vessels that Yara and Anduril are scaling into civilian and military maritime use, the Spot quadrupeds that Boston Dynamics has now deployed to oil rigs, talent shows, and presidential residences — every one of these platforms exists because someone at MIT, Carnegie Mellon, Stanford, Berkeley, ETH Zurich, Oregon State, the Florida Institute for Human and Machine Cognition, or one of roughly thirty other research-grade university robotics laboratories spent a decade building the precursor system that the commercial product is descended from. The university research robotics ecosystem is the technological R&D pipeline that the rest of the cluster has been spending. The K-12 robotics competitions feed students into that pipeline. The pipeline feeds commercial products into every other domain in the cluster. The middle layer — the university lab and the NASA mission and the national research facility — is where the technology actually gets invented.

The Agility-Cassie-Digit lineage and the university-to-commercial pipeline

The clearest example of the pipeline in 2026 is Agility Robotics, the company that built the bipedal humanoid platform Digit that is currently being commercially deployed in pilot programs at Amazon warehouses, GXO Logistics facilities, and Spanx distribution centers. Digit is, in lineage terms, the direct commercial descendant of Cassie — the open-source, ostrich-legged dynamic locomotion research platform that Agility’s founders developed at Oregon State University‘s Dynamic Robotics Laboratory under principal investigator Jonathan Hurst. Cassie spun out of OSU in 2017. The intervening eight years have been a series of progressively more capable Cassie iterations, the introduction of upper limbs and a head to create Digit, the build-out of a Salem, Oregon factory capable of producing Digit at volume, and the recent commercial scaling that has put the platform on the floor of working warehouses. The DARPA Robotics Challenge of 2013-2015 — the program that gave rise to the modern humanoid-robot industry — was won by Team KAIST’s DRC-Hubo with MIT‘s Atlas variant, IHMC‘s Atlas, Tartan Rescue’s CHIMP from Carnegie Mellon, and several others in the top finisher list. Every one of those teams was a university or research-institute team operating under DARPA funding. The companies those teams seeded — Boston Dynamics, Apptronik, Figure (whose founder Brett Adcock came from the IHMC orbit), Sanctuary, 1X — have raised, collectively, north of $25 billion in venture capital across the subsequent decade. The research-to-commercial path is a fifteen-to-twenty-year lag, and it is the dominant path by which serious humanoid robotics has reached the market.

The same pipeline runs through every other major commercial robotics platform the cluster has documented. Boston Dynamics Spot is the commercial descendant of BigDog, the DARPA-funded quadruped that Marc Raibert’s group at the MIT Leg Laboratory began developing in the early 2000s before spinning out as Boston Dynamics in 1992 and continuing the work through Google’s ownership (2013-2017), SoftBank’s ownership (2017-2020), and Hyundai’s ownership (2020-present). Anduril Dive-LD descends from the AUV-research work conducted at the Woods Hole Oceanographic Institution and MIT’s Hatx Lab over twenty years. Saildrone descends from Richard Jenkins’s land-yacht and ocean-yacht engineering experiments, ultimately influenced by the Naval Postgraduate School’s autonomous-sailing research. Zipline‘s autonomous-fixed-wing-medical-delivery platform descends from the same family of autonomous-flight research that the Stanford GPS Lab, MIT’s Aerospace Controls Laboratory, and Berkeley’s Center for Information Technology Research in the Interest of Society spent the 2000s and 2010s building. The naming conventions change. The institutional sponsorship changes. The underlying claim — that university research labs are the upstream source of commercial robotics — does not.

The Mars rover program as the boundary case

NASA’s Mars exploration program — and its analogues at the Chinese CNSA, the European Space Agency, the Indian Space Research Organisation, and the Japanese JAXA — represent the extreme upper bound of what scientific research robotics is capable of producing. The current operational fleet on Mars consists of NASA’s Perseverance rover (landed February 18, 2021), NASA’s Curiosity rover (landed August 6, 2012 and still operating), and the Chinese Zhurong rover (landed May 14, 2021, dormant since 2022). The retired Ingenuity helicopter still sits at Valinor Hills, transmitting weather telemetry weekly. Perseverance is, in any quantitative sense, the most sophisticated autonomous robotic platform humans have ever sent to another world: a 2,260-pound, plutonium-238-thermoelectric-generator-powered, six-wheeled rover carrying a 7-foot robotic arm with five degrees of freedom, a 24-tube sample caching system, a 23-camera imaging suite, a ground-penetrating radar, an organic-molecule detector, an X-ray fluorescence spectrometer, and the in-situ resource utilization experiment MOXIE that successfully demonstrated the production of breathable oxygen from Martian atmospheric CO₂. Perseverance is, depending on how you allocate ground-segment costs across the mission lifetime, in the range of a $3 billion robot.

The Mars Sample Return mission — the multi-decade program intended to physically retrieve the samples Perseverance has been collecting and return them to Earth for laboratory analysis — has been the most consequential scientific research robotics program restructuring of the 2020s. The original baseline architecture, finalized in 2022, depended on a NASA-built Sample Retrieval Lander, a Mars Ascent Vehicle, an ESA-built Earth Return Orbiter, and a sample-handling system that combined to roughly $11 billion in lifecycle cost. By mid-2024, that estimate had grown to $11-13 billion with a return-to-Earth date no earlier than 2040. NASA Administrator Bill Nelson initiated a major program review in April 2024 to consider alternative architectures, and in early 2025 NASA selected dual study contracts with Lockheed Martin and a SpaceX-Rocket Lab consortium to evaluate lower-cost commercial alternatives. The resulting program is, as of 2026, fundamentally restructured around a more aggressive commercial-launch baseline and a shorter timeline, with the original Sample Retrieval Lander concept effectively cancelled. The largest scientific research robotics program the United States has ever attempted is being rebuilt, mid-flight, around a fundamentally different commercial-industrial logic than the one that produced Perseverance — which is the same commercial-industrial logic that the rest of this cluster has been documenting in adjacent domains.

The Berkeley A-Lab and the self-driving-laboratory wave

The closest analogue inside terrestrial science to the Mars rover’s autonomous-scientific-decision-making is the self-driving laboratory — the integrated robotic-and-AI platform that designs experiments, runs them, interprets the results, and decides what to do next, without human intervention in the loop. The most publicized example in 2023-2026 was the A-Lab at Lawrence Berkeley National Laboratory, built by Gerbrand Ceder‘s materials science group at UC Berkeley in collaboration with Yan Zeng, Kristin Persson, and a Google DeepMind team. The A-Lab was published in Nature in late November 2023 with a claim that, over 17 days of continuous autonomous operation, the system had performed roughly 21 experiments per day and produced 41 novel inorganic compounds out of an attempted 58 — a 71% success rate, with the inputs drawn from the Materials Project database and DeepMind’s GNoME (Graph Networks for Materials Exploration) catalog of computationally predicted candidate materials. The paper was, in the materials-discovery community, treated as the closest thing to a fully autonomous scientific discovery system anyone had built.

The Nature publication was followed, in early December 2023, by a detailed critique from Robert Palgrave, a materials chemist at University College London, who argued in a widely-circulated X thread that the A-Lab’s automated phase-identification system had misclassified most of the supposed novel compounds and that, on closer inspection of the X-ray diffraction data, the system had not in fact synthesized any new materials. Ceder responded on LinkedIn in late December 2023, defending the underlying methodology while conceding that “a human can perform a higher-quality refinement on these samples.” A more formal critique by Palgrave and collaborators followed in 2024. As of late 2025, the consensus position across the materials-discovery community — captured in a December 2025 MIT Technology Review feature titled “AI materials discovery now needs to move into the real world” — was that despite the A-Lab’s documented technical capability to operate autonomously around the clock, no convincing breakthrough discovery had emerged from any of the major self-driving lab projects, and Ceder himself had taken a sabbatical from Berkeley to become Chief Science Officer at Radical AI, a New York City materials-discovery startup setting up its own self-driving labs in commercial space. The most ambitious autonomous-scientific-discovery program built to date had, in the operational reading, produced infrastructure but not yet results. The cluster’s recurring observation that the publicity has outrun the deliverables applies — perhaps more sharply in this domain than anywhere else.

The broader self-driving-laboratory ecosystem extends well beyond Berkeley. Alán Aspuru-Guzik‘s group at the University of Toronto operates one of the longest-running autonomous chemistry platforms. The Acceleration Consortium at Toronto, launched in 2023 with a $200 million CFREF Canadian federal grant, is building a network of self-driving labs across Canadian universities focused on clean energy materials. Carnegie Mellon‘s autonomous chemistry group, led by Lee Cronin at Glasgow with his “chemputer” platform, operates a different architecture aimed at autonomous synthesis of pharmaceutical molecules. The MIT Bayesian Reaction Optimization group has produced a series of autonomous optimization platforms used in industrial chemistry pilot lines. Opentrons sells open-source pipetting robots into research labs at price points that have made bench automation accessible to academic groups that could not previously afford Hamilton, Tecan, or Beckman Coulter systems. The combined deployed footprint of self-driving labs and bench-automation platforms across academic research is, by 2026, somewhere in the tens of thousands of installations — small compared to the warehouse-robot installed base, but compounding rapidly and concentrated in the highest-value scientific output per dollar spent.

The university lab as humanoid-robot proving ground

The most operationally consequential deployment of commercial robotics into university research environments is the use of Boston Dynamics Spot, ANYbotics ANYmal, and Agility Robotics Cassie/Digit as standard research platforms across roughly two hundred robotics laboratories worldwide. MIT CSAIL operates multiple Spots and a Boston Dynamics Atlas research platform. Stanford’s robotics group operates Spots, an ANYmal, and a fleet of Skydio drones. Carnegie Mellon’s Robotics Institute operates Spots, ANYmal C, an Atlas, a custom CHIMP humanoid descendant, and one of the largest TurtleBot fleets in the United States. ETH Zurich’s Robotic Systems Lab — the academic group that originally spun out ANYbotics — operates ANYmal extensively for legged-locomotion research and is one of the most prolific publishers of legged-robot autonomy research in the world. UC Berkeley’s Robot Learning Lab under Sergey Levine operates a mix of commercial platforms and custom prototypes. Caltech’s Center for Autonomous Systems and Technologies operates Spots and a fleet of custom drones. The Florida Institute for Human and Machine Cognition (IHMC) continues to operate the modified Atlas platforms it inherited from the DARPA Robotics Challenge era. The University of Tokyo, Tokyo Institute of Technology, KAIST, Tsinghua, Shanghai Jiao Tong, the Italian Institute of Technology, EPFL Lausanne, the University of Edinburgh, and TU Delft round out the global research-grade university robotics ecosystem.

The structural argument that makes this deployment matter is that the same Spot platform that reads gauges on BP’s Mad Dog and the same ANYmal that operates on the Petrobras P-71 platform are, fundamentally, refined versions of research platforms that were running open-source autonomy stacks in graduate student labs five to ten years earlier. The commercial product cycle in robotics is, structurally, slower and more research-dependent than the commercial product cycle in software. The next generation of commercial robot — Figure 03, Apptronik Apollo 2, Boston Dynamics Atlas’s hydraulic-to-electric transition, Agility Digit’s next-generation manipulation upgrades — depends in measurable part on what’s happening in graduate-level robotics research right now. The reader who has spent the cluster reading about policing drones and autonomous mining trucks and Trajekt Arc baseball-pitching robots is, in this section, looking at the upstream R&D environment those products are being incrementally drawn out of, by research groups whose annual budgets are typically less than the cost of a single mid-tier commercial humanoid robot.

ROS, TurtleBot, and the open-source infrastructure

The software substrate that makes the entire university-research-robotics ecosystem function is ROS — the Robot Operating System — originally developed at Stanford and Willow Garage in the late 2000s, transferred to the Open Source Robotics Foundation in 2012, and now maintained by Open Robotics, the foundation’s commercial arm that was acquired by Apex AI in late 2022. ROS is the de facto operating system for academic robotics — virtually every research-grade university robotics platform in the world either runs ROS natively or includes a ROS-compatibility layer. The TurtleBot — the open-source mobile robot platform originally designed at Willow Garage in 2010 and now in its TurtleBot 4 generation — is the global standard educational and research mobile-robot platform, with installed-base estimates in the tens of thousands across university labs, community college programs, and high-end K-12 STEM facilities. Clearpath Robotics‘ Husky and Jackal unmanned ground vehicles are the heavier-duty commercial alternatives. Universal Robots’ UR3, UR5, UR10, and UR16 collaborative robotic arms — manufactured in Odense, Denmark, and now owned by Teradyne — are the standard commercial-bench robotic arm in research labs across roughly seventy countries. Franka Emika‘s Panda is the research-grade German alternative. Kinova Robotics‘ Gen3 ultra-lightweight arm is the standard for robotics research requiring portability or human-collaborative operation.

The economic structure of this ecosystem is that the open-source foundation (ROS, TurtleBot, Gazebo simulation) creates the substrate on top of which commercial platforms (UR, Franka, Kinova, Clearpath, Boston Dynamics, ANYbotics, Agility) compete. The substrate is sustained by university research output. The commercial platforms are sold back into the same university labs whose research produced the substrate. The same NVIDIA Jetson and NVIDIA Orin compute platforms that run Disney’s BDX droid and the Skydio X10 also run the typical TurtleBot or Husky deployment. The same lithium-ion battery chemistry, the same rare-earth permanent magnets, and the same semiconductor supply chain that the rest of the cluster has documented show up across the entire research-robotics hardware stack. The component supply chains are convergent. The application domains are divergent.

The drone side: wildlife, environmental, and atmospheric science

The drone-side of scientific research robotics produces a different category of work. The NOAA Hurricane Hunter Reconnaissance Squadron uses unmanned Black Swift S0 and Coyote drones launched into the eyewalls of hurricanes to measure central pressure, wind shear, and storm structure at altitudes and conditions where crewed Lockheed WP-3D Orion aircraft cannot safely operate. The British Antarctic Survey and the U.S. Antarctic Program routinely deploy fixed-wing drones to map ice-shelf calving fronts, count penguin colonies (the Penguin Watch project’s drone fleet has surveyed hundreds of millions of square meters of Antarctic coastline since 2017), and monitor seal populations on remote South Georgia and South Orkney islands. The University of Hawaii flies drones into active volcanic vents at Kīlauea, Mauna Loa, and the Halemaʻumaʻu caldera for plume sampling and lava-flow mapping under conditions that would kill a crewed aircraft. The National Park Service flies drones across Yellowstone for geyser-system monitoring and across Glacier National Park for ice-mass-balance measurements that historically required helicopter-borne teams at orders-of-magnitude higher cost.

In oceanographic research, Saildrone Voyager units are now standard equipment for NOAA fisheries assessments, hurricane-eye intercepts (the first-ever in-storm video from inside a Category 4 hurricane was captured by a Saildrone in Hurricane Sam in 2021), and Arctic methane-flux measurements. REMUS AUVs from HII (formerly Hydroid) are the standard 3-meter-class autonomous underwater vehicle for academic oceanography. WHOI’s Nereus hybrid ROV reached the Mariana Trench in 2009 and operated at full ocean depth before its loss in 2014. The MBARI Mesobot operates at midwater depths tracking individual zooplankton over hour-long observation windows that crewed submersibles cannot sustain. The combined research-grade autonomous-vehicle fleet across all U.S. academic oceanography programs is, by NOAA estimates, in the low thousands of units across the surface, midwater, and deep-ocean tiers — and is the underlying R&D pipeline that produced the maritime-defense-robotics market that Anduril and Saildrone are scaling into U.S. Navy and Allied operational use.

What 2026 looks like in research and university robotics

In 2026, Boston Dynamics Spot, ANYbotics ANYmal, and Agility Robotics Cassie are operating in approximately two hundred university research robotics laboratories worldwide. NASA’s Perseverance rover continues to operate at Jezero Crater, having collected 27 sample tubes that are now slated for a restructured Mars Sample Return program scheduled to return them no earlier than the late 2030s. The Berkeley A-Lab continues to operate, with the Nature-paper controversy unresolved and the underlying autonomous-experimentation infrastructure being adopted by Radical AI, the Acceleration Consortium at Toronto, and a handful of pharmaceutical-industry sites. ROS — the Robot Operating System — runs on virtually every university research-grade robotics platform on Earth. TurtleBot, Husky, Jackal, UR5, Franka Panda, and Kinova Gen3 remain the standard commercial-research hardware. NOAA Hurricane Hunter drones, British Antarctic Survey penguin-counting drones, Saildrone Voyagers in the Arctic and Pacific, REMUS AUVs in academic oceanography, and the long tail of specialized scientific drones across volcanic monitoring, wildlife research, and atmospheric sampling continue to produce the published data that fills the journals. The DARPA Robotics Challenge cohort of 2013-2015 continues to produce the commercial humanoid-robot industry that the cluster’s first post documented. The K-12 FIRST and VEX teams are continuing to feed into the universities. The universities are continuing to feed into the commercial robotics industry. The Mars sample return program is being rebuilt around commercial launch economics.

The research robots in this cluster do something different than every other category of robot the cluster has documented. They are not optimizing margins on warehouse picking. They are not patrolling oil platforms or hospital corridors. They are not delivering blood to remote villages or dropping water on California wildfires. They are not pitching baseballs or dancing on talent shows. The research robots in 2026 are demonstrating, in graduate student labs and DOE-funded national laboratories and NASA mission ops centers and Antarctic field stations, what robots will be capable of in five to fifteen years. Ingenuity proved Mars helicopters are possible. The A-Lab proved autonomous materials synthesis is possible, with the open question of whether it can be made into reliable discovery still being argued in peer-reviewed comments and X threads and conference panels. Cassie proved that bipedal robots can run, and Digit is now stacking warehouse totes. ROS proved that an open-source operating system could become the universal substrate of an entire industry, the same way Linux did for the server market a generation earlier. Saildrone Voyager proved that a 23-foot solar-and-wind-powered sailing vessel can spend twelve months at sea without human intervention and bring back hurricane data the U.S. Navy and NOAA cannot get any other way. The thing every one of these platforms shares is that they were built in research environments where the immediate operational ROI was not the point — the point was to demonstrate that the thing could be done. Once it could be done, the rest of the cluster picked it up and built the product.

The most consequential robots in human history — the ones on Mars, the ones at the bottom of the Mariana Trench, the ones that mapped the genome, the ones that imaged the first black hole, the ones that demonstrated autonomous flight on another world — were all built in scientific research environments by graduate students, postdocs, and mission-systems engineers whose names are mostly not in the press. The 2026 cohort of research robotics is the cohort whose work will, fifteen years from now, populate the rest of this cluster with the next generation of commercial deployments. Ingenuity does not fly anymore. The 17 kilometers it covered, the 72 missions it completed, and the proof-of-concept it delivered for atmospheric flight on another planet are the cluster’s clearest possible example of what research robotics is for. The rest of the robotics industry, in 2026, is built on top of the foundation of work that platforms like Ingenuity, like Perseverance, like Cassie, like ANYmal, like Saildrone, and like the A-Lab were built to test. The graduate students assembling the next generation of those platforms in basement labs at MIT and Stanford and CMU and ETH Zurich and Tokyo and KAIST and Tsinghua are, this spring, doing the upstream work that the rest of American workforce development and the rest of the global robotics market is, in the cluster’s running thesis, structurally dependent on.
Classroom, School and Childhood Education Robotics in 2026: The Talent Pipeline That Built Every Other Robot in the Cluster

On Saturday morning, January 10, 2026, in approximately 3,700 high school auditoriums, gymnasiums, and engineering classrooms across thirty countries, students aged 14 to 18 watched a synchronized video feed from FIRST headquarters in Manchester, New Hampshire, as the FIRST Robotics Competition revealed the year’s game. The format has not changed in any significant way since the program’s first season in 1992. Each team receives an identical “kit of parts” containing motors, wheels, sensors, a control system, and structural materials. The teams have exactly six weeks to design, build, program, and test a 125-pound robot capable of playing a game whose rules they learned that morning for the first time. The 2026 season — called REBUILT, presented by Haas, with kickoff sponsored by Qualcomm — runs through eight weeks of regional qualifying tournaments in March and early April, with the world championship scheduled for April 29 through May 2 at the George R. Brown Convention Center in Houston. The teams that win their regional events advance to Houston. The teams that win in Houston earn the right to be called world champions of an event that, in any quantitative sense, is the largest annual gathering of secondary-school engineers anywhere on Earth.

FIRST — For Inspiration and Recognition of Science and Technology — was founded in 1989 by the inventor Dean Kamen, the same Dean Kamen who invented the Segway, the AutoSyringe insulin pump, the iBOT motorized wheelchair, the home dialysis system that became the foundation of DEKA Research, and roughly five hundred other patented devices. Kamen, with assistance from MIT professor emeritus Woodie Flowers, designed the FIRST format around a single thesis: that American secondary education was failing to produce the engineers the country would need to build the next century’s economy, that the failure was structural rather than incidental, and that the solution was not better textbooks but a national robotics competition that would make engineering feel like a varsity sport. The 1992 inaugural FIRST competition had 28 teams in a New Hampshire high school gym. The 2024 season had 3,701 active FRC teams in 30 countries and regions, with parallel competitions — FIRST LEGO League for elementary and middle school (35,140 teams at last published count), FIRST Tech Challenge for older middle and high school, FIRST LEGO League Explore for the youngest students — adding another 50,000-plus teams to the global FIRST footprint. The competition has, in operational terms, become the most successful workforce-development program in modern American technology history.

This is the domain where the cluster’s running thesis about robotics meets the question of where the engineers who actually build the robots come from. The humanoid robots at Figure AI and Apptronik, the Skydio drones operating Drone-as-First-Responder programs for 1,500 American police departments, the autonomous haul trucks running across Rio Tinto’s Pilbara mine, the Boston Dynamics Spot platforms patrolling BP’s Mad Dog offshore platform, the Trajekt Arc pitching robots in nineteen MLB clubhouses, and the Sikorsky-Rain autonomous Black Hawk dropping water on California wildfires were, almost without exception, designed by engineers who built their first robot in a high school workshop for a FIRST or VEX competition. The talent pipeline worked exactly the way Kamen designed it. The 2026 cohort building robots in those 3,700 gymnasiums is the cohort that will, in roughly a decade, be building the next generation of every robotic platform the rest of this cluster has been describing.

VEX Robotics and the parallel competition ecosystem

The competing platform — and, by raw team count, the larger of the two — is VEX Robotics, founded in 2007 by Tony Norman and Bob Mimlitch at Innovation First International in Greenville, Texas, with the VEX V5 Robotics Competition (V5RC) as its high school flagship and the VEX University Robotics Competition (VURC) for colleges. VEX operates more than 20,000 registered teams across more than 50 countries, with the 2025-2026 game called Push Back and the world championship held annually at the Kay Bailey Hutchison Convention Center in Dallas. The 2026 V5RC High School World Championship was won by team 1028A, “WASHED,” with Excellence Award honors going to 9181C, “C-Channel.” The team names — “WASHED,” “Exothermic Burnout,” “Cyber Spacers,” “Iron Panthers,” “The Cheesy Poofs” (FIRST team 254, the all-time leader with five championship titles) — are the kind of distinctively-teenaged branding decisions that have, in retrospect, become a load-bearing cultural feature of the entire ecosystem. The kids who name their robot “Exothermic Burnout” are the kids who go on to win the DARPA Robotics Challenge ten years later.

The VEX program differs from FIRST in two structurally important ways. First, the build budget is dramatically lower — a competitive VEX team can field a robot for under $1,500 in parts, where a FIRST FRC robot routinely costs $5,000 to $15,000 between kit-of-parts components, custom machining, and travel — which makes VEX the dominant program in lower-income school districts and in countries where corporate sponsorship is thinner. Second, the VEX season is shorter and the games are smaller-scale, which puts the design and iteration cycle on a tighter clock and produces a different style of engineer — faster, scrappier, less reliant on adult mentorship. RoboCup, founded in 1996 with the explicit goal of fielding a fully autonomous robot soccer team capable of defeating the human World Cup champions by 2050, sits adjacent to both programs as the global research-grade competition, drawing university teams from MIT, Carnegie Mellon, Tokyo Institute of Technology, ETH Zurich, and a long roster of others. BEST Robotics, founded in 1993 in Sherwood, Texas, occupies the middle ground with a more limited parts budget and a creative-problem-solving emphasis. The combined global footprint of secondary-school robotics competition — FIRST + VEX + BEST + RoboCup juniors + a long tail of national and regional programs — is, by team count, well over 80,000 teams and somewhere north of a million students participating annually as of 2026.

LEGO Education, Sphero, Wonder Workshop, and the elementary-school tier

Below the competition tier is the classroom-product tier — the physical robotics kits and platforms designed for elementary and early middle school instruction. LEGO Education dominates this market with the SPIKE Prime kit (launched in 2019, designed for grades 6-8 around a programmable Bluetooth hub and the LEGO Technic part system), the SPIKE Essential kit for grades 1-5, and the LEGO Mindstorms EV3 that defined the category from 2013 until LEGO discontinued the Mindstorms consumer line in October 2022 to consolidate development around the Education-branded products. The discontinuation announcement was, in the robotics-education community, treated roughly the same way the discontinuation of a popular automobile model would be treated by the car enthusiast community — a piece of cultural infrastructure being shut down for reasons that were essentially commercial rather than pedagogical. The SPIKE Prime that replaced it is a more sophisticated product but does not have the consumer-retail availability or the brand recognition of the Mindstorms line it succeeded.

Sphero, founded in 2010 in Boulder, Colorado, sells the BOLT programmable robotic ball, the indi elementary classroom robot, and the RVR programmable rover, with classroom-pack configurations sold to school districts at volume pricing. Wonder Workshop, founded in 2012 in Sunnyvale, California, sells the Dash, Dot, and Cue robots for elementary classroom use, with a curriculum tied to the Blockly visual programming language used in over 20,000 American elementary schools. Ozobot sells line-following educational robots that respond to color-coded markers drawn on paper. Makeblock in Shenzhen sells the mBot line into the Chinese, European, and increasingly American elementary classroom market, part of the broader Chinese ed-tech-and-hardware export effort that has, in education robotics as in agricultural drones and consumer drones, captured large global market share through aggressive price competition. The hardware in every one of these platforms depends on the same semiconductor supply chain, the same rare-earth permanent magnets in the motors, and the same lithium-ion battery chemistry as every commercial robotics platform the rest of the cluster has documented — a fact that the K-12 robotics-education market mostly does not advertise, but that creates a generation of students who, by the time they enter the workforce, have grown up assuming that the components in their classroom kits and the components in industrial robots are the same components. KIBO by KinderLab Robotics in Waltham, Massachusetts, sells a screen-free programmable robot specifically designed for ages 4-7, marketed against the early-childhood-screen-time concerns that have been intensifying across pediatric medicine since roughly 2018. The combined U.S. classroom-robotics market — across all elementary and middle school products — is, depending on definition, somewhere in the $400 million to $700 million range annually, growing at roughly 8-12% per year, and dominated almost entirely by physical hardware rather than software.

The Khanmigo curve and the AI tutor wave

In parallel — and, in the 2024-2026 ed-tech conversation, almost entirely crowding out the physical-robotics-in-schools story — is the AI-tutor wave. Khan Academy launched Khanmigo, its GPT-4-powered AI tutor and teaching-assistant platform, on March 14, 2023, the same day OpenAI publicly released GPT-4. Khan Academy was one of OpenAI’s earliest external partners for GPT-4 access. The product is a custom-prompted GPT-4 wrapper, designed to engage students in Socratic-style questioning rather than to supply direct answers. The initial rollout was a paid pilot for donors and selected schools. By the 2023-2024 school year, Khanmigo had roughly 40,000 K-12 student users. By 2024-2025, that number had jumped to 700,000 — what Khan Academy’s chief learning officer Kristen DiCerbo publicly described as “the biggest one-year jump that I have seen in terms of adoption of an education technology” in twenty years of ed-tech work. The projected 2025-2026 user base is over one million students. New Hampshire became the first U.S. state to sign a statewide Khanmigo partnership in June 2024, with the agreement extended through 2025-2026 at no cost to the state under Khan Academy’s nonprofit-pricing model.

Khanmigo is not a robot. It is a chat interface running on top of OpenAI’s API, with a curriculum-aware prompt structure that Khan Academy has been refining for three years. The product is the closest thing the ed-tech industry has produced to what venture capital has been promising since the early 2010s — a personalized one-on-one tutor available 24 hours a day for every student in the world. The pedagogical evidence for the product is mixed. The 60 Minutes feature in December 2024, the New Hampshire statewide deployment, the Microsoft partnership that made Khanmigo’s teacher tools free globally in 34+ languages, and the expansion into school systems in India, Brazil, and the Philippines have made it the highest-profile AI-in-education product in the world. Whether it actually improves measured student outcomes remains, as of 2026, an open empirical question that the published clinical-trial-grade evidence has not yet conclusively answered — though the user-growth curve has been steep enough that the question may be answered by adoption rather than by research.

The South Korean AI textbook disaster

The cautionary case in 2025-2026 was South Korea. Under former president Yoon Suk Yeol, the South Korean Ministry of Education spent roughly 1.2 trillion won — approximately $830 million U.S. — to develop and deploy AI Digital Textbooks (AIDT) in elementary, middle, and high school classrooms beginning in March 2025. Publishers invested another 800 billion won developing 76 approved AIDT titles. The program was launched as mandatory for English, mathematics, and computer science instruction in grades 3-4 of elementary school, first-year middle school, and first-year high school. The Ministry of Education trained 1,200 “digital tutors” and committed an additional $43.2 million to install monitoring systems in 6,000 primary and secondary schools.

The program collapsed in approximately four months. The AI textbooks failed to recognize numbers handwritten by students, flagged correct answers as wrong, and produced what users described as “nonsensical responses.” Teachers reported their workload had increased rather than decreased. Parents organized — 56,505 signatures on a petition opposing the rollout in May and June 2024 alone, with 86% of polled parents and teachers opposing the program by the time it launched. The Korean Teachers and Education Workers Union and the civic group Political Mamas sued the Minister of Education in November 2024 for abuse of authority. By October 2025, over half of the 4,095 schools that had signed onto the program had opted out. After Yoon’s impeachment and removal from office in 2024, his successor revoked the official “textbook” status of the AIDT, reclassifying them as “supplementary materials” — leaving publishers who had invested in the platform without the legal mandate they had developed against. As of late 2025, AIDT adoption sits at roughly 30% across approved subjects, mostly in schools whose principals chose to continue using the materials as optional supplements.

The South Korean disaster is the cleanest published case of “government rollout of AI in schools” failing at national scale. Compare it to the LAUSD “Ed” AI chatbot disaster — the Los Angeles Unified School District‘s March 2024 launch of an AI student-support chatbot that collapsed within months after the contracted vendor, AllHere Education, filed for bankruptcy and the system was found to have serious privacy and functionality issues. The pattern is consistent across both: an ambitious top-down deployment of AI-powered software into the K-12 classroom, justified by promises of personalized learning and teacher-workload reduction, that runs into operational problems and political backlash within months of launch. By contrast, FIRST Robotics — operating with no government mandate, no top-down deployment, and no software-vendor lock-in — has grown its team count every year since the COVID-19 pandemic-driven decline of 2020-2021, with the 2026 season on track to exceed pre-pandemic team counts and the Houston championship venue expanding accordingly.

The talent pipeline argument

The structural argument that makes the robotics-in-education category cluster-relevant is the workforce pipeline. Boston Dynamics‘ senior engineering staff includes Atlas program leads who participated in FIRST as high school students in the mid-2000s, and whose graduate-school research projects involved earlier-generation Spot platforms acquired by university robotics labs. Skydio‘s founding team came out of MIT and Stanford robotics laboratories whose member rosters were dominated by FIRST and VEX alumni. Figure AI‘s engineering leadership includes alumni of the DARPA Robotics Challenge teams from MIT, Carnegie Mellon, and Florida Institute for Human and Machine Cognition (IHMC) — all of whom had, in the previous decade, been products of the FIRST competition pipeline. Anduril‘s autonomy and drone teams, Zipline‘s aerospace engineering staff, Saildrone‘s naval-architecture group, Trajekt Sports‘ mechanical-engineering team that built the 1,200-pound MLB pitching robot, BRINC Drones‘ robotics group that built the LEMUR 2 indoor tactical drone — the talent supply across the entire commercial robotics industry, by every available company-history reconstruction, draws disproportionately from the FIRST and VEX competition pipeline.

The reason this matters is that the secondary effects of the FIRST and VEX investment are structurally enormous and almost entirely uncaptured by the program’s nominal mission statement. The kid who joins the FIRST team in tenth grade and learns to weld an aluminum frame, machine a custom gearbox, debug a Java control loop, and write a competition strategy briefing is the kid who, fifteen years later, is leading the autonomy team at Tesla Optimus or designing the next generation of Husqvarna’s robotic lawnmower fleet or running the Skydio engineering organization. The capital cost of that training pathway — measured per resulting professional engineer — is, by every available comparison, dramatically lower than the capital cost of the equivalent university engineering program, and dramatically lower than the capital cost of the failed AI-textbook deployments. The investment in physical robotics in K-12 returns to the broader robotics industry in the form of engineers who can do the work. The investment in software-based AI tutoring returns to the AI industry in the form of, in many cases, contracts that get cancelled within a year of signing.

The Dean Kamen original thesis, in 1989, was that the United States needed to produce more engineers and that the production process should look more like the Friday night football game in a Texas high school and less like an AP Chemistry class. The intervening thirty-seven years of FIRST competition have produced, by FIRST’s own published alumni data, somewhere in the range of two to three million participants who went on to STEM careers at a rate roughly double the U.S. high school baseline. The single largest population of robotics engineers in 2026 — across warehouse automation, agricultural drones, maritime autonomy, policing drones, healthcare robots, Japanese eldercare platforms, and the humanoid-robot demo cycle that has consumed the cluster’s attention since the cluster opened — was produced, in measurable part, by a Manchester, New Hampshire nonprofit and a Greenville, Texas company that have been quietly running the same competition format for somewhere between 19 and 37 years.

Drones in the K-12 classroom

The drone side of the K-12 ecosystem is smaller and more recently formed but follows the same workforce-pipeline logic. DJI Education sells the Tello EDU and RoboMaster TT programmable drone kits into elementary and middle school classrooms with curricula tied to Scratch and Python programming. Skydio has, through its public-safety and government channels, supplied small numbers of drones to U.S. high schools running drone-pilot certification programs aligned to the FAA Part 107 Remote Pilot Certificate. The Aerial Sports League runs high-school drone-racing competitions in California, Texas, and a handful of other states. The TSA-Approved Drones in Schools program is exploring K-12 deployment of small drones for STEM curriculum integration. By comparison to the FIRST/VEX physical-robotics ecosystem, the K-12 drone education footprint is still small — measured in the low thousands of participating schools rather than the tens of thousands — but the growth curve since 2022 has been steep, and the pipeline-feeding effect into the commercial drone industry is, by analogy, expected to produce a similar workforce dividend over the next decade.

The companion robot, the autism-therapy robot, and the small categories

Three smaller categories complete the K-12 robotics picture. The first is the autism-therapy companion robot, with LuxAI‘s QTrobot (designed for ages 4-14 with autism spectrum disorder), SoftBank‘s Pepper (used in autism therapy programs across Japan, France, and the U.K.), and Embodied‘s Moxie (a small tabletop social robot for child emotional development, founded by former Jibo CEO Paolo Pirjanian) as the leading platforms. The clinical evidence base for therapy-robot interventions is modest but growing, with published trials documenting improvements in attention, social engagement, and routine compliance for children using the platforms in structured therapeutic contexts. The second is the classroom telepresence robot, with VGo (acquired by Vecna Robotics in 2015) and Double Robotics providing remote-attendance platforms for medically homebound students — a category that saw a dramatic but temporary expansion during COVID-19 and has since settled back to a smaller specialized market. The third is the library-and-makerspace robot — Sphero RVRs, LEGO SPIKE kits, and 3D-printer-and-robotics combo packs that increasingly populate the maker spaces being built in renovated school libraries across the United States, a quiet infrastructure investment that has, in operational terms, replaced the school-library budget as the single largest line item for elementary-school technology purchasing.

What 2026 looks like in robotics education

In 2026, FIRST has 3,700-plus active FRC teams, 35,000-plus FIRST LEGO League teams, and a combined K-12 footprint approaching half a million participants annually across all program tiers. VEX Robotics has 20,000-plus registered competition teams in 50-plus countries. LEGO Education’s SPIKE platform is deployed in roughly 60,000 American elementary and middle school classrooms. Sphero, Wonder Workshop, Ozobot, Makeblock, and KIBO collectively serve another 100,000-plus American elementary classrooms. The Khanmigo K-12 user base has crossed one million students, distributed across hundreds of school districts in the United States and pilot programs in India, Brazil, and the Philippines. The South Korean AI Digital Textbook program is in the process of being unwound, with adoption sitting at roughly 30% under voluntary use and the underlying publishers absorbing the losses from the cancelled mandate. The Los Angeles Unified School District’s “Ed” chatbot has been functionally retired. The 2026 FIRST Championship at Houston’s George R. Brown Convention Center is scheduled for April 29 through May 2. The 2026 VEX Worlds in Dallas is scheduled for early May. The kids who will design the next generation of every robot in the rest of this cluster are, this spring, finishing six-week build seasons in roughly four thousand school workshops across thirty countries.

The robotics-in-education category is the cluster’s most upstream domain. The robots in classrooms are not the robots in mines, ports, hospitals, oil rigs, theme parks, or wildfire zones. They are the robots that the future builders of those other robots first encountered as teenagers in a gymnasium in January. The investment in FIRST and VEX over thirty-seven years has produced a workforce dividend that the rest of the robotics industry has been spending, with extraordinary results, across every domain the cluster has documented. The investment in AI-powered classroom software, in the same period, has produced mostly headlines — and, in South Korea’s case, an $830 million government-funded headline that lasted four months. The pattern across the cluster has been consistent. The robots that work are deployed where the work is real, the engineering is grounded, and the workforce has been trained for thirty years to make machines do hard things in physical environments that punish error. The kids in those four thousand workshops in January are building the single most important raw material the robotics industry will consume over the next thirty years. Dean Kamen built that pipeline in 1989 on a hunch about Texas high school football. The hunch turned out to be the most consequential bet in the history of American workforce development, and the rest of this cluster is, in 2026, what that bet finally cashed out as.
Maritime Robotics in 2026: Ports, Offshore, and the Salmon Farms Running Themselves

The robots that move 90 percent of global trade do not have legs, faces, or names. They are 100-foot ship-to-shore cranes that pick 50-ton steel containers off the deck of a vessel the length of four American football fields and place them on autonomous battery-electric trucks that drive themselves to stacks of other containers managed by autonomous stacking cranes operating in a yard the size of a small city. The whole choreography happens at the Port of Rotterdam with 10 to 15 humans per shift moving 14 million containers per year, at the Port of Singapore’s new Tuas Mega Port — designed to handle 36 million TEUs at full build-out by the 2040s — with electric autonomous guided vehicles that emit zero carbon and run on an AI-orchestrated fleet management system, and at Yangshan Deepwater Port in Shanghai with what is now generally regarded as the most heavily automated container terminal on Earth. The global market for automated container terminal equipment was $11.3 billion in 2025 and is projected to reach $22.4 billion by 2035. Roughly 80 percent of the ship-to-shore cranes currently operating at U.S. ports were built by a single Chinese state-owned manufacturer. None of this gets the coverage that a humanoid robot doing a backflip gets. All of it is doing significantly more economic work.

The same applies underwater, where Anduril Industries’ Quonset Point, Rhode Island factory is being scaled up to produce up to 200 Dive-LD autonomous underwater vehicles per year alongside larger Dive-XL mothership platforms in Sydney, Australia, with the U.S. Navy and the Royal Australian Navy committing multi-year contracts to the buildout. And on the surface, where the Pentagon’s Replicator initiative is buying autonomous surface vessels and undersea vehicles by the hundreds rather than the dozens for the first time in the post-WWII history of American naval procurement. And in the salmon pens off the coast of Norway, where 1.4 billion farmed Atlantic salmon are being monitored, fed, deloused, and harvested by an emerging stack of underwater robotics that includes Tidal — an Alphabet X spin-off — and AKVA Group’s submerged-cage Nautilus system that has cut sea lice treatments by an order of magnitude.

Three subdomains. Three different sets of acronyms. One underlying observation: the maritime sector is where the most consequential robotics deployment is happening, and the companies that get the headlines for humanoid robots are not, with rare exception, the same companies that are doing the work.

The container port and the ZPMC problem

A modern container terminal is built around three families of equipment, all of which can be automated and most of which already are. The ship-to-shore (STS) crane lifts the box off the vessel onto the dock — at the largest ports, the new generation of “double-trolley” STS cranes can lift two containers simultaneously and operate without a human in the cab. The automated stacking crane (ASC) moves containers within the storage yard. And the automated guided vehicle (AGV) — a battery-electric, GPS-guided platform with no driver — transports the container between the STS crane and the ASC stack. Singapore’s Tuas Mega Port runs the whole sequence on AI-orchestrated AGV fleets with 50 percent lower carbon emissions than diesel terminal trucks. Rotterdam’s Maasvlakte II has been operating versions of this stack since 2014. Yangshan Phase IV in Shanghai opened in 2017 as the world’s largest automated container terminal at the time. Long Beach Container Terminal in California runs partial automation — quay cranes still manned, but yard operations largely automated — and has been the source of running disputes with the International Longshore and Warehouse Union (ILWU) that has shaped American labor politics around the technology in ways that the Asian operators have not had to navigate.

The dominant supplier of the heavy equipment is Shanghai Zhenhua Heavy Industries — ZPMC — a state-owned Chinese manufacturer controlled by China Communications Construction Company. ZPMC built around 80 percent of the ship-to-shore cranes currently operating at U.S. ports and roughly 70 percent of the cranes operating worldwide. A single ZPMC STS crane costs $10 to $15 million and the company can underprice every Western competitor because it does not face the same profit pressure as a publicly traded engineering firm. Finland’s Konecranes is the only meaningful Western alternative for new STS crane purchases. There is no domestic American manufacturer of ship-to-shore cranes at all. The U.S. Navy, the FBI, and the Cybersecurity and Infrastructure Security Agency have publicly stated that they consider ZPMC equipment a potential vector for cyber-intrusion into U.S. port infrastructure, with cellular modems and other communications hardware embedded in the cranes that the Pentagon has reportedly compared to “a Trojan horse.” The Biden administration imposed a 25 percent tariff on Chinese-made STS cranes in 2024. On October 14, 2025, the second Trump administration’s U.S. Trade Representative finalized an additional 100 percent tariff on the same equipment, effective November 9, 2025, with a carve-out for cranes contracted before April 17, 2025 and delivered before April 18, 2027. ZPMC publicly warned in May 2025 that the tariff would “cripple U.S. ports.”

This is the civilian-military-fusion model the Chinese state has refined for decades applied to the equipment that loads every container of every product that arrives in the United States by sea. ZPMC’s parent, China Communications Construction Company, is sanctioned by multiple U.S. federal agencies for work on artificial islands in the South China Sea. The same company sells cranes to U.S. ports. The cranes contain electronic components — modems, sensors, controllers — that originate in the same Chinese supply chain the U.S. is simultaneously trying to decouple from in semiconductors and in critical metals like gallium and germanium, and that the United Front Work Department playbook has been steering through civilian commercial channels into U.S. critical infrastructure for years.

The American port industry is, structurally, asking the federal government a hard question with no good answer: if Chinese cranes are a national security risk, and no American manufacturer makes them, and the European alternatives can’t scale fast enough, and the next decade of container vessel growth requires new cranes — where exactly do the cranes come from? Konecranes is expanding. American startups are talking about entering the market. The Port of Virginia is asking for a 12-month phase-in. The cranes are still being ordered from ZPMC under the pre-April 17 carve-out. The infrastructure dependency turns out to be the kind of legacy commitment that is much easier to enter than to exit — and the alternative, which is to build a domestic crane industry from scratch over 5 to 10 years, costs more money than anybody is currently willing to commit and produces no political payoff before the next election.

In the meantime, the rest of the world is moving in the opposite direction. China’s three-step plan to dominate the maritime sector targets becoming a global innovation hub by 2025 and the world’s leading maritime power by 2035. Seven of the top 10 busiest container ports on Earth are Chinese. Yangshan Port is roughly 50 percent more productive than Rotterdam on every productivity metric available. The Port of Tanjung Pelepas in Malaysia signed a deal in February 2025 to buy 58 ZPMC rubber-tired gantry cranes. Beyond the United States, the question is not whether to build with ZPMC. The question is how fast.

The undersea drone economy

Above water, the defense robotics buildout has happened in the open — Boston Dynamics, Tesla, Figure, the humanoid roundup. Below water, the same buildout has happened more quietly, on contract numbers that dwarf the consumer-facing announcements. The U.S. Navy operates Unmanned Undersea Vehicle Squadron 1 (UUVRON-1) out of Keyport, Washington — one of two Navy squadrons whose entire mission is to develop, test, and deploy underwater drones. On April 5, 2025, Anduril delivered the first Dive-LD to UUVRON-1: a 6-meter-long autonomous undersea vehicle capable of operating at depths up to 6,000 meters, with 10-day endurance, modular payloads, and a 3D-printed hull design that allows production rates the legacy submarine industry cannot match. In August 2024, the Pentagon selected the Dive-LD as part of the second tranche of the Replicator initiative — a program designed to mass-produce autonomous systems in the thousands to deter Chinese military expansion in the Indo-Pacific. In March 2026, the Navy selected Anduril’s larger Dive-XL for the CAMP program, which positions the platform as an underwater “mothership” that can carry and launch smaller undersea drones, including the company’s torpedo-launchable Copperhead — a sub-class platform unveiled at Sea Air Space 2025 in two variants (Copperhead-100 and Copperhead-500) that fit inside Dive-XL’s payload bay the way a fighter jet’s missiles fit inside its weapons bay.

The undersea robotics industry is, in operational terms, a generation older than the humanoid robotics industry. Commercial remotely operated vehicles (ROVs) have been doing oil-and-gas inspection at depths up to 4,000 meters since the 1980s. Companies like Oceaneering, Subsea7, Saipem, and TechnipFMC have been operating ROV fleets for decades. The 2025-2026 shift is from human-piloted ROVs tethered to a surface vessel to genuinely autonomous autonomous underwater vehicles (AUVs) that can operate untethered for days at a time, executing pre-programmed missions and adapting to conditions on the fly without continuous human oversight. The technology that makes that shift possible is the same family of perception and autonomy software that has enabled the autonomous weapons buildout above water, the same machine-vision pipelines that are enabling autonomous spray drones on farms and autonomous Spot platforms in defense procurement. The water makes the engineering harder. The hardware is more expensive. The acoustic communications channels are vastly narrower than the radio spectrum available to surface drones. But the fundamental capability — perceive environment, plan action, execute, repeat without supervision — is the same.

Saildrone, the Alameda-based company founded by Richard Jenkins, is the surface-vessel equivalent. Saildrone’s autonomous Voyager USV — a 33-foot wind-and-solar-powered sailing platform — has been used by the U.S. Navy, NOAA, and a dozen other government and commercial customers for missions ranging from hurricane data collection to maritime domain awareness in the Pacific. The company is now adapting the platform into a long-endurance anti-submarine warfare platform that can patrol contested ocean for months at a time at a fraction of the cost of a frigate. Anduril’s Ghost Shark, an extra-large AUV produced under contract with the Royal Australian Navy, delivered its first operational platform in 2025 and is being scaled up at a Sydney production facility. Turkey’s Sefine ULAQ USV is in service with the Turkish Navy. The Ukrainian Navy’s Magura V5 and Sea Baby unmanned surface vessels have, in the course of the war in the Black Sea, sunk or damaged more Russian naval tonnage than any other category of weapon since 2022 — using vessels that cost roughly $250,000 each, packed with explosives, and steered toward Russian warships by operators sitting in Kyiv. The cost-asymmetry logic that drove loitering munitions in Ukraine has now translated, in nearly identical form, to the maritime domain.

What this means in 2026 is that the undersea environment, which for most of human history has been the domain of nation-state navies operating expensive manned submarines, is becoming a contested space where companies like Anduril, Saildrone, and L3Harris can produce hundreds of autonomous vessels per year at unit costs that any country with a reasonable defense budget can afford to buy in volume. AUKUS — the Australia-UK-US security partnership announced in 2021 — explicitly identifies autonomous undersea systems as a Pillar Two technology priority, alongside the nuclear-powered submarines at the center of the agreement. The Pacific deterrent posture the United States is building against the Chinese navy depends, increasingly, not on the dwindling number of attack submarines the U.S. Navy can deploy, but on the rapidly increasing number of autonomous underwater drones it can manufacture in Quonset Point and Sydney.

The salmon farm running itself

Norway produces more than 1.4 million tons of farmed Atlantic salmon annually, which is roughly half of the global supply and represents 73 percent of Norway’s seafood export revenue. The industry is concentrated along the Norwegian coastline in tens of thousands of submerged net pens, each one holding up to 200,000 salmon. The single largest operational problem facing the industry is sea lice — a parasitic copepod that attaches to the skin of farmed salmon, causes welfare problems, reduces growth rates, and triggers regulatory penalties when infestation thresholds are exceeded. Manual sea-lice removal — done by lifting the fish out of the water, hot-water bathing them, or applying chemicals — is stressful for the salmon, hazardous for the workers, and expensive for the producer. The Norwegian salmon industry spends an estimated $700 million per year on sea-lice mitigation.

Underwater robotics has become the operational backbone of how that mitigation now happens. Stingray Marine Solutions, a Norwegian startup, operates underwater drones equipped with computer vision and surgical diode lasers that detect a sea louse on a passing salmon and kill the parasite with a 100-millisecond laser pulse — without lifting the fish, without applying chemicals, without manual intervention. Tidal, a spin-off from Google X — the same Alphabet moonshot factory that produced the autonomous-driving company Waymo and the geothermal start-up Dandelion — launched Tidal Lice Control at the AquaNor 2025 trade show in Trondheim. The system is an AI-driven autonomous platform that operates inside salmon pens around the clock, detecting and neutralizing lice without manual handling. AKVA Group, the largest publicly traded aquaculture technology company, has commercialized Nautilus — a deep-farming solution where the salmon are kept in submerged net pens with a surface air dome for swim-bladder access, and where data from six commercial sites shows 0.6 delousing operations per pen versus 6.1 at conventional surface sites. Remora Robotics of Stavanger has built fully autonomous net-cleaning and inspection robots that operate continuously inside the pens, preventing biofouling without the high-pressure cleaning that stresses the fish.

The underwater drone fleet inside a 2026-vintage Norwegian salmon operation is, in scale terms, larger than the surface drone fleet at the average mid-sized agricultural operation in the American Midwest. Deep Trekker, a Canadian ROV manufacturer, has hundreds of small inspection ROVs operating in Norwegian fish farms doing everything from sea-lice counting to net inspection to monitoring lumpfish — a cleaner-fish species that aquaculture operators stock in pens specifically to eat sea lice off the salmon as a biological alternative to chemical or laser intervention. Aquaai, a San Diego startup, has deployed robotic fish — actual computer-vision-equipped artificial salmon that swim alongside the real ones — to provide non-intrusive monitoring inside cages with up to 188,000 individuals. The same fundamental observation from Japanese elder-care robotics applies here in mirror-image form: the robots that work in the field are the ones that solve a discrete, well-defined problem (sea lice detection, net cleaning, individual fish health monitoring) — not the ones that try to replace the entire labor pool with a single general-purpose machine.

What the Norwegian aquaculture industry has built over the last decade is, in operational terms, the closed-loop precision agriculture pattern applied underwater: scout robots gather data, AI processes the data into prescriptions, action robots execute the prescriptions, results are measured by the scout robots in the next cycle. The same architecture that runs autonomous DJI Agras spray drones over Brazilian soybean fields is running underwater laser-equipped sea-lice killers in Norwegian fjords, and the productivity gains are comparable. The 2025 Norwegian parliament debates on biomass limits and welfare measures are happening inside a regulatory environment that explicitly assumes a high-automation production model — the alternative, which is a return to chemical delousing and manual net cleaning, is no longer politically or environmentally viable, which means the industry is locked into the robotics path whether individual operators prefer it or not.

The autonomous ship that almost works

One last piece. The Yara Birkeland, an 80-meter, 120-TEU, fully battery-powered container ship operated by the Norwegian agricultural-chemical company Yara International, has been operating commercially in Norwegian coastal waters since 2022. It is the world’s first commercial-operation autonomous container vessel. It can self-dock, self-cross, and self-discharge. It eliminates an estimated 40,000 diesel truck journeys per year. And it still operates with a crew of three onboard — recently reduced from a larger initial complement — supervising the autonomous systems for regulatory reasons that have nothing to do with whether the autonomy actually works. The two-year autonomy trial period that was originally supposed to conclude in late 2024 has been extended. The shore-based remote operations center in Horten is fully built and operational. The vessel is functionally autonomous and operationally crewed. The same gap between technical capability and regulatory permission that holds back drone delivery in 2026 holds back autonomous shipping, in the same shape and roughly the same proportions, with the same set of insurers, regulators, and labor unions deciding the pace of the rollout.

The Mayflower Autonomous Ship, developed by ProMare and IBM and capable of crossing the Atlantic without a crew, made its maiden voyage in 2022. The Sea Hunter, DARPA’s anti-submarine warfare USV, has been in continuous Navy operation since 2018. Hyundai Heavy Industries, Maersk, Wallenius Wilhelmsen, and most of the world’s major shipowners have active autonomous-vessel research programs. Nothing about the technology is the bottleneck. The bottleneck is the same regulatory, insurance, and labor-relations question that defines every other domain where robots are entering the workforce: who carries the liability when something goes wrong, who pays the unemployment claim when the workers are displaced, and which government agency owns the certification authority that determines whether the autonomous system is allowed to operate.

What 2026 actually looks like across the maritime sector

A container ship leaves Yangshan in Shanghai loaded by a ZPMC-built crane onto a vessel managed by a Chinese-owned shipping line, sails the Pacific monitored by a fleet of Saildrone Voyager USVs collecting maritime domain awareness data for the U.S. Navy and the Anduril Dive-LD autonomous undersea vehicles operating below the surface in patterns that the People’s Liberation Army Navy cannot fully observe, arrives at the Port of Long Beach where a partially automated terminal moves the containers off the vessel using cranes built by ZPMC, transferred to autonomous battery-electric AGVs that run on the same kind of copper-dense electric drivetrain that powers every other large-scale electrification project on the planet — and is then loaded onto the same diesel trucks that have been carrying containers out of American ports since the 1950s, because the last-mile logistics of the surface freight network is the part of the chain where automation is happening slowest, in the same operational pattern visible at every robotics-adoption frontier. Up the coast in Norway, in a salmon pen that holds 200,000 individuals, an autonomous Stingray laser drone is killing sea lice at a rate of one parasite per 100 milliseconds while a Remora Robotics net cleaner does its scheduled biofouling sweep and a Deep Trekker ROV runs an opportunistic inspection of the cage perimeter — and the entire operation is overseen by two technicians sitting in a control room in Trondheim, supervising 17 sea-cage installations across the Norwegian coast simultaneously, in a working pattern that resembles the supervisory model that healthcare robots have begun to enable in American hospitals and that no humanoid robot manufacturer has yet operationalized at scale.

The robots in maritime do not look like robots. They look like cranes, like submersibles, like sailing platforms, like fish. They do not perform on stage. They move 90 percent of global trade, they patrol the ocean floor under contracts the public never sees, and they keep half the world’s farmed salmon alive long enough to reach a refrigerator. They are the deployment side of the same industry whose humanoid demos generate the headlines, and they are doing the work the headlines describe — quietly, in volume, in a working economy that depends on them more completely each year, and that, in 2026, is being reshaped by a U.S.-China trade fight over port cranes, a Pentagon scaleup of undersea drone manufacturing, and a Norwegian aquaculture industry that has built the world’s most heavily automated food production system on the back of a copepod the size of a grain of rice that nobody outside the salmon business has ever heard of.