News / Updates / Blog:

  • Marc Rich and Glencore: The Fugitive Who Built the World’s Largest Commodity Trader

    In 1983, a federal grand jury in New York indicted Marc Rich on 65 criminal counts — income tax evasion, wire fraud, racketeering, and trading with Iran during the hostage crisis in violation of U.S. sanctions. The potential sentence exceeded 300 years. It was the largest tax evasion case in American history at the time, prosecuted by a young federal attorney named Rudolph Giuliani. Rich learned of the indictment, flew to Switzerland, and never returned. He stayed on the FBI’s Ten Most Wanted Fugitives list for years, narrowly escaping capture in Finland, Germany, Britain, and Jamaica. He didn’t even return for his daughter’s funeral in 1996. And on January 20, 2001 — his last day in office, among 140 pardons and commutations — President Bill Clinton gave Marc Rich a full and unconditional pardon. The New York Times called it “a shocking abuse of presidential power.” Jimmy Carter said the pardon was “disgraceful.” The company Rich built from his exile in Zug, Switzerland, had by then become the largest commodity trading firm on earth. It is still operating. Its name is Glencore.

    The invention of modern commodity trading

    Rich was born Marcell David Reich in Antwerp in 1934. His Jewish family fled the Nazis through Vichy France, Spain, and Portugal, arriving in the United States aboard the liner Serpa Pinto. He dropped out of college in New York and went to work in the mailroom at Philipp Brothers, then the world’s dominant metals trading house. He was a prodigy. By his mid-twenties he was making deals across Europe; by his thirties he was Phibro’s top producer. In 1973, during the OPEC oil embargo, Rich figured out how to bypass the cartel’s ban on sales to the United States, buying cargoes from one company and reselling them to another on a short-term basis. He essentially invented the crude oil spot market — the system of buying and selling individual cargoes of oil outside of long-term contracts that defines global oil trading to this day.

    Furious over his compensation, Rich left Phibro in 1974 with his partner Pincus “Pinky” Green and founded Marc Rich + Co. AG in Zug, Switzerland. The choice of Zug was not incidental. Swiss law at the time drew a distinction between tax evasion (a civil matter) and tax fraud (a criminal matter). Switzerland interpreted its neutrality doctrine so strictly that it declined to enforce many international trade embargoes. And Zug’s tax rates were among the lowest in Europe. Rich had found the jurisdiction that would let him trade with anyone, pay minimal taxes on the proceeds, and resist extradition from the country whose laws he was breaking.

    The sanctions portfolio

    Rich traded with everyone the United States told its citizens not to trade with, and he was explicit about why. “You can’t run a business based on sympathies,” he told his biographer Daniel Ammann. “Otherwise our business would be hampered.” The client list reads like a sanctions compliance officer’s nightmare: Iran during and after the hostage crisis, apartheid South Africa, Cuba under Castro, Libya under Gaddafi, Ceaușescu’s Romania, Pinochet’s Chile, Sandinista Nicaragua, Marxist Angola.

    The Iran-South Africa oil pipeline was his masterpiece of sanctions arbitrage. Iran, post-revolution, was under U.S. embargo and couldn’t easily sell its crude. South Africa, under UN sanctions for apartheid, couldn’t easily buy oil. Both were desperate — Iran to sell, South Africa to buy. Rich positioned himself as the only trader willing to bridge the two pariah states, extracting enormous margins from both sides because neither had alternative counterparties. The structural logic was identical to what made BCCI valuable to its clients: when legitimate channels are closed, the middleman who operates outside the law captures the entire spread. Rich’s companies earned an estimated $2 billion from these trades alone.

    Rich also served as an asset for Israeli intelligence. He reluctantly acknowledged in interviews with Ammann that he had assisted the Mossad, a claim confirmed by a former Israeli intelligence officer. Rich financed Mossad operations and supplied Israel with strategic quantities of Iranian oil through a secret pipeline arrangement. This dual role — private businessman and intelligence asset — would become critical to his pardon. When the pardon effort began, it was coordinated by Avner Azulay, a former high-ranking Mossad agent who had been running Rich’s philanthropic foundations in Israel since 1993. Azulay persuaded Rich’s ex-wife Denise to appeal directly to Clinton. He also enlisted Israeli Prime Minister Ehud Barak to call Clinton on Rich’s behalf.

    The pardon

    The mechanics of the pardon are the part that connects Rich to the Shadowcraft thesis. Denise Rich — who had divorced Marc in 1996 — donated $450,000 to the Clinton Presidential Library Foundation and over $100,000 to Hillary Clinton’s Senate campaign. Leonard Garment, Nixon’s former special counsel, represented Rich. Scooter Libby — later convicted in the Plame affair, later pardoned by Trump — served as Rich’s attorney until 2000. The lobbying campaign deployed former intelligence officials, Israeli heads of state, and major Democratic donors in a coordinated effort to secure clemency for a man on the FBI’s Most Wanted list.

    Clinton’s defense was that the charges were better adjudicated through civil rather than criminal procedure. Eric Holder, then deputy attorney general, later testified that if he had known all the facts, he would not have recommended the pardon. Congress launched a bipartisan investigation. The episode became shorthand for the proposition that wealth and political connections can purchase outcomes the justice system was designed to prevent — a proposition that Rich’s entire career had already demonstrated through commodity markets rather than courtrooms.

    What Rich built — and what it became

    In 1993, Rich sold Marc Rich + Co. to his management team. They renamed it Glencore. Under CEO Ivan Glasenberg — who had joined the firm in 1984 and worked his way up through the South African coal trading desk — Glencore became the world’s largest commodity trading company and one of the largest publicly traded companies on earth. It went public in 2011 and merged with mining giant Xstrata in 2013, creating Glencore Xstrata (later just Glencore), a vertically integrated behemoth that trades and mines copper, cobalt, zinc, nickel, coal, oil, and agricultural commodities across every continent.

    Rich died in 2013 in Switzerland. He was 78. He was buried in Israel. But the corporate culture he built — the willingness to trade with sanctioned regimes, the use of intermediaries and shell structures to obscure transactions, the treatment of bribery as an operating expense — survived him. In May 2022, Glencore pleaded guilty in the United States to one count of conspiracy to violate the Foreign Corrupt Practices Act. The company admitted to paying more than $100 million in bribes to government officials in Nigeria, Cameroon, Ivory Coast, Equatorial Guinea, Brazil, Venezuela, and the Democratic Republic of Congo between 2007 and 2018. Separately, it pleaded guilty to commodity price manipulation. The combined penalties across U.S., UK, and Brazilian proceedings exceeded $1.1 billion. In August 2024, Swiss authorities convicted Glencore of “inadequate organisation” leading to corrupt mine deals in the DRC, imposing an additional $152 million penalty.

    The DRC case is the one that illustrates the mechanism. Glencore used Dan Gertler — an Israeli businessman and mining middleman now on the U.S. sanctions list — to negotiate mining deals with the government of then-president Joseph Kabila. When Glencore acquired a majority stake in Kamoto Copper Company, one of the world’s largest copper-cobalt mines, Gertler negotiated a $440 million discount on the signing bonus. Glencore paid $140 million instead of $585 million. The difference — money that should have gone to the Congolese state — disappeared into the gap between what Glencore paid and what the asset was worth. Gertler continues to receive tens of thousands of dollars daily in royalty payments from these mines. Glencore’s $180 million settlement with the DRC covers “all present and future claims” from 2007 to 2018, buying permanent immunity from further prosecution for a fraction of the revenue the mines generate in a single year.

    The accounting line item for bribes in Glencore’s 1990s-era books was labeled “useful expenses.” That phrase tells you everything about the continuity between Marc Rich + Co. and the company that inherited its culture.

    What it means

    Marc Rich invented modern commodity trading. He also invented the modern template for sanctions evasion as a business model — positioning yourself in the jurisdictional gap between the countries imposing sanctions and the countries subject to them, using Swiss neutrality and corporate opacity as infrastructure, and treating the legal risk as a cost of doing business rather than a constraint on behavior. Russia’s shadow fleet runs on the same structural logic Rich pioneered in the 1970s: find the parties who can’t trade through legitimate channels, insert yourself as the intermediary, extract the premium, and structure the operation through jurisdictions that won’t enforce the sanctions. The shell company architectures are the same. The flag-of-convenience registries are the same. The willingness to treat enforcement risk as a pricing input rather than a moral constraint is the same.

    Rich died wealthy, pardoned, and free. Glencore paid $1.1 billion in fines and kept operating. The Congolese communities that lost hundreds of millions in mining revenue have received a fraction in opaque settlements. The system Rich built — where the intermediary captures the value and the source country absorbs the loss — is the system the Shadowcraft course is designed to make visible. Not because it’s secret. Because it’s legal enough to survive prosecution and profitable enough that the fines are a line item.

    We cover Marc Rich alongside BCCI, Crypto AG, Wagner Group, and 20 other case studies of covert institutional power across our Shadowcraft course — where “useful expenses” is the two-word summary of how the world actually works.

  • The Ngogo Chimpanzee War: The First Documented Civil War in a Non-Human Species

    On the last full day of his life, a chimpanzee named Basie woke at dawn in a tree nest he’d built from branches and leaves, surrounded by other chimps dozing in their own nests, as he’d done nearly every day for 36 years in the Kibale National Park rainforest in Uganda. He spent an ordinary day swinging between trees and eating figs. As daylight faded, a patrol of about 13 adult chimpanzees from the opposing faction arrived. Three surrounded him. He jumped from a tree. Ten piled on him on the ground, biting him. Basie’s killers were chimpanzees he had grown up with — individuals he had groomed, traveled with, and defended territory alongside for decades. His death in 2019 was the second casualty in what researchers now call the Ngogo chimpanzee civil war, an eight-year conflict that has killed at least 28 chimpanzees, including 19 infants, and that a study published in Science on April 9, 2026, has documented in detail that primatologists say is unprecedented.

    What happened

    The Ngogo chimpanzee community was the largest known group of wild chimpanzees on earth — approximately 200 individuals living in relative cohesion in Kibale National Park for at least 20 years under continuous scientific observation since 1995. Chimpanzee communities typically number around 50. Ngogo was four times that. The group operated through a fission-fusion social structure — small parties formed and dissolved throughout the day as individuals moved around the territory foraging and socializing, but everyone belonged to the same community, shared the same territory, and collectively defended it against neighboring groups. Within the community, social relationships clustered around two primary neighborhoods that researchers named the Central and Western groups, but the boundary was porous. Chimps changed which cluster they associated with. Males groomed partners from both groups. Females mated across the divide. Key individuals — socially connected males who maintained relationships in both clusters — served as bridges holding the community together.

    Then those bridges collapsed. Several of the bridging males died from disease. A new alpha male rose to power, shifting the community’s political center of gravity. A respiratory disease outbreak further destabilized social networks. By approximately 2015, chimps in the Western and Central clusters began avoiding each other. The avoidance hardened into separation. By 2018, the division was permanent — two distinct communities with separate territories, separate social hierarchies, and no remaining social bonds between them.

    What followed was not a border skirmish between strangers. It was coordinated lethal violence between former companions. The Western faction — numerically smaller, starting at about 76 individuals — launched targeted raids into Central territory. Groups of adult males would patrol into enemy territory, locate isolated individuals, and attack with overwhelming numbers. The violence was graphic: sustained group assaults, biting, mutilation. From 2021, the Western raiders began targeting and killing infants — a pattern that primatologists associate with territorial expansion, as infanticide eliminates the offspring of rivals and can make females sexually receptive sooner.

    The Western faction’s campaign has been described as a “one-sided rout.” Their numbers grew from 76 to 108 over the course of the conflict. The Central faction suffered a stepwise decline. John Mitani, a professor emeritus at the University of Michigan who had been studying the Ngogo chimps for two decades when the violence started, told NBC News he is concerned the Central group is “doomed.” The war is ongoing. The 2026 Science paper covers data through 2024, but lead author Aaron Sandel of the University of Texas at Austin confirmed that further attacks have occurred in 2025 and 2026.

    Why it matters

    This is only the second documented case of a chimpanzee community splitting and going to war with itself. The first was the Gombe Chimpanzee War of the 1970s, observed by Jane Goodall in Tanzania, where a community called the Kasakela fissioned and the splinter group (the Kahama) was systematically hunted and destroyed over four years. The Gombe war was groundbreaking but limited by the observational methods available in the 1970s. The Ngogo study benefits from 30 years of continuous demographic data, 24 years of systematic behavioral observations, a decade of GPS tracking, and structured social network analysis — a dataset that Gombe never had. Genetic evidence suggests that permanent community fissions in chimpanzees are extraordinarily rare, occurring roughly once every 500 years. Researchers have now documented two in 50 years of field primatology, which either means the estimate is wrong or scientists have been spectacularly unlucky — or lucky, depending on your perspective.

    The social network data is what makes the Ngogo study new. The 2026 Science paper mapped the social ties between individuals across the entire community for years before, during, and after the split. What they found is that the division didn’t happen along genetic lines, or resource boundaries, or any clear ecological gradient. It happened along social network lines. When the bridging individuals who maintained connections between the two clusters died or were removed, the network fragmented — and fragmentation preceded violence by approximately three years. The chimps didn’t fight and then separate. They separated and then fought. Avoidance came first. Identity formation second. Lethal violence third.

    Aaron Sandel told BBC Science Focus that the study provides “a window into the chimpanzee mind that’s really rare” — the transition from friend to enemy, visible in behavioral data over a decade. The implication for understanding human conflict is the part that’s getting the most attention. In humans, collective violence is typically explained by cultural differences — ethnicity, religion, language, ideology — that bind groups together and generate hostility toward outsiders. But the Ngogo chimps had no cultural markers distinguishing the two factions. They spoke the same calls, ate the same food, lived in the same forest, and had mated with each other for years. The split wasn’t driven by what made them different. It was driven by the decay of what had kept them connected.

    Sandel’s conclusion is pointed: if chimpanzee civil wars emerge from the breakdown of interpersonal relationships rather than from intergroup differences, then human peace interventions that focus on cultural diplomacy — learning about the other side’s traditions, bridging ideological divides — may be missing the more fundamental mechanism. “What we have to do is maintain interpersonal relationships,” Sandel told Scientific American. “If we can reunite — even in the face of conflict — then I think that’s a recipe for maintaining peace.” Liran Samuni of the German Primate Center, who was not involved in the study, noted that even before the split, the Ngogo community was “one of the chimpanzee communities that was most violent in terms of encroaching on neighbors” — they had previously killed at least 21 chimpanzees from other groups and expanded into their territory. The civil war is new. The violence isn’t.

    The Gombe parallel

    Anne Pusey, who conducted fieldwork at Gombe until 1975 during the beginning of that war, told the Washington Post that the circumstances preceding both conflicts were “similar and shocking”: a shortage of mating-age females, the death of socially central older males, a change in alpha male, and disease. In both cases, social bonds that had been stable for years degraded rapidly once key connective individuals were removed from the network. Joseph Feldblum, an evolutionary anthropologist who has studied the Gombe data, said the Ngogo findings validate the earlier observations: “This sort of behavior, while rare, is part of the natural course of chimpanzee behavior.”

    The baboon politics research on coalition formation and dominance hierarchies, the chimpanzee tool use literature documenting cultural transmission across generations, and the dolphin signature whistle work demonstrating individual identity in non-human social systems all converge on the same insight the Ngogo war makes visceral: complex social cognition isn’t an abstract capacity. It’s the infrastructure that holds societies together — and when the infrastructure fails, the consequences in chimpanzee communities look disturbingly like the consequences in human ones. Former friends become lethal enemies not because something changed about who they are, but because the relationships that made them “us” instead of “them” stopped being maintained.

    We cover the Ngogo war alongside mirror neurons, corvid intelligence, animal deception, and 20 other investigations into what animal minds reveal about the architecture of social life across our Animal Culture & Knowledge course — where the question isn’t whether animals have societies but what happens when those societies break.

  • Warehouse Robots in 2026: Amazon, Ocado, and the Automation of Logistics

    Amazon’s fulfillment network processed its fastest delivery year ever in 2025. Not because it hired more people — though it did — but because it deployed more robots alongside them. The company operates the largest fleet of warehouse robots on earth, hundreds of thousands of autonomous mobile units that carry shelving pods to human pickers, sort packages into delivery routes, and move pallets across facilities the size of small airports. Amazon doesn’t disclose exact fleet numbers anymore, but the scale is visible in the architecture: its newest fulfillment centers are designed around robot traffic patterns the way cities are designed around roads. The humans work in stations. The robots own the floor.

    Amazon isn’t alone, but it is the scale benchmark that defines what everyone else is chasing. The warehouse robotics market was valued at $12.85 billion in 2024, growing at approximately 20 percent annually, projected to reach $53.5 billion by 2032. More than 1.3 million robotics-as-a-service deployments are expected by 2026 worldwide. The labor economics are the accelerant: warehouse work is physically demanding, turnover is high, hiring is difficult during peak seasons, and a robot that costs a subscription fee per month doesn’t call in sick, doesn’t require benefits, and doesn’t quit after three weeks because the guy at the facility across the highway is paying a dollar more per hour.

    The three models

    Warehouse automation has settled into three competing architectures, each with a different answer to the same question: what do you automate first?

    Amazon’s model is goods-to-person. Autonomous mobile robots carry entire shelving units — pods stacked with inventory — to stationary human pickers who stand at ergonomic workstations and pull items for orders. The human does the picking (the dexterous, judgment-intensive part). The robot does the walking (the repetitive, time-consuming part). This architecture eliminates the single biggest inefficiency in a traditional warehouse: human travel time between picks. Amazon’s earliest Kiva robots (acquired in 2012 for $775 million) established the template. The current generation is faster, denser, and integrated with AI-driven inventory placement that positions high-velocity items closer to pick stations.

    Ocado’s model is the grid. The British online grocery company built automated fulfillment centers where thousands of small robots — roughly the size of a washing machine — run on a grid structure above a three-dimensional storage matrix of crates. The robots navigate the grid, lower grabbers into the matrix to retrieve specific crates, bring them to pick stations, and return the crates to optimized storage positions. Ocado’s system processes an online grocery order in minutes — picking 50 items from a catalog of 50,000 SKUs — with picking accuracy and throughput that human-only operations can’t match at the same speed. Ocado acquired 6 River Systems from Shopify in 2023, adding collaborative mobile robots (the Chuck platform) to its portfolio. The company now licenses its technology to grocery retailers globally — Kroger in the U.S., Coles in Australia, and multiple European chains — making Ocado as much a robotics company as a grocery company.

    Symbotic’s model is full-facility automation. The company builds systems where fleets of small, fast-moving robots operate across racking structures, handling storage, retrieval, case picking, sorting, and pallet building with minimal human involvement. Symbotic’s robots pick individual cases from pallets five times faster than human workers and build shipping pallets optimized by AI to reduce product damage by 30 percent during transit. In January 2026, Symbotic acquired Walmart’s Advanced Systems and Robotics division for $200 million, with Walmart simultaneously investing $520 million in Symbotic to deploy AI-powered robotics across its distribution network. The deal gave Symbotic a backlog exceeding $5 billion. Target, Albertsons, and Kroger are also clients. Symbotic’s bet is that the entire distribution center — not just the pick station — should be automated end to end.

    The unsolved problem

    The constraint that limits all three models is the same one, and it’s not the hardware. MIT researchers working with Symbotic published findings identifying fleet coordination — not individual robot capability — as the primary bottleneck in large-scale automated fulfillment. A warehouse running hundreds of robots simultaneously is fundamentally a traffic management problem. Robots converge on the same aisles, queue at the same pick stations, and create congestion patterns that mirror rush-hour traffic on a highway. MIT’s approach uses machine learning to predict where congestion is forming and reroute robots before the jam develops, achieving roughly 25 percent higher throughput than traditional routing algorithms in tests modeled on real e-commerce warehouse layouts.

    The picking problem is the second constraint. Robots are excellent at moving things — carrying pods, transporting crates, navigating grids. They are mediocre at picking things — reaching into a bin of mixed items, identifying the correct product, grasping it without damage, and placing it in an order container. This is the task that still requires human hands in Amazon’s model and Ocado’s model. Companies like Covariant (founded by AI researchers from UC Berkeley and OpenAI) have built robotic arms with AI “brains” that can handle items the system has never seen before — critical for e-commerce where SKU variety is enormous and changes daily. Covariant’s robots are deployed across apparel, pharmaceuticals, electronics, and third-party logistics operations, reducing picking errors and eliminating the training time that new human pickers require. But fully autonomous picking across the full range of warehouse SKUs — from a bag of chips to a bottle of shampoo to a pair of shoes — remains unsolved at the reliability level that replaces a human picker entirely.

    What 2026 actually looks like

    The industry lesson from 2025, according to multiple retrospective analyses, was that reliability beats novelty. Companies that deployed mature, well-integrated systems outperformed those that adopted cutting-edge robotics that required constant troubleshooting. The differentiator in 2026 isn’t the robot itself — it’s the orchestration layer: the software that coordinates humans, robots, and existing equipment into a coherent workflow. Facilities designed around how humans and robots intersect, rather than around any single automation investment, consistently produce better results.

    Robotics-as-a-service has changed the economics of adoption. Instead of $1 million-plus capital expenditure for a full automation installation, warehouses can subscribe to robot fleets on monthly contracts — adding units during peak season and scaling down after. The model mirrors SaaS in enterprise software: lower barrier to entry, predictable costs, continuous upgrades, and the ability to trial automation without betting the facility on it. ABI Research projects this subscription model will account for the majority of new warehouse robot deployments by 2026.

    The workforce story is more nuanced than “robots replace workers.” Automation has shifted what warehouse workers do rather than eliminating their jobs entirely. Repetitive transport tasks disappear. Quality control, robotics coordination, maintenance, kitting, and exception handling increase. The humanoid robot race — with companies like Agility Robotics deploying Digit humanoid robots in Amazon facilities — represents the next attempt to automate the physical tasks that current robots can’t perform: walking through unstructured environments, manipulating objects of arbitrary shape and weight, and working in spaces designed for human bodies rather than robot grids.

    Whether the end state is a warehouse with zero humans or a warehouse where humans do fundamentally different work than they do today is the question that Symbotic’s $5 billion backlog and Amazon’s fleet expansion are simultaneously trying to answer. The honest assessment from 2025 is that full automation of a general-purpose warehouse is further away than the vendor pitches suggest, closer than the skeptics claim, and happening in specific, high-volume, low-variability operations — grocery distribution, pallet handling, sortation — faster than almost anyone predicted five years ago.

    We cover warehouse automation alongside robot dogs, drone delivery, Japan’s elder care robots, and the full spectrum of machines entering the physical world across our Humanoid Robots & Drones course — where the question isn’t whether robots will change logistics but whether logistics is the rehearsal for everything else.

  • Desalination in 2026: The Technology, the Cost Curve, and the Gulf States Betting Their Future On It

    Ninety percent of Kuwait’s drinking water comes from desalination. In Oman, it’s 86 percent. In Saudi Arabia, 70 percent. The Gulf Cooperation Council countries account for roughly 60 percent of global desalination capacity, producing 40 percent of the world’s desalinated water from over 400 plants. These countries didn’t adopt desalination because it was cheap or elegant. They adopted it because the alternative was having no water. The global desalination market was valued at $20.3 billion in 2023 and is projected to reach $44.6 billion by 2032, growing at a compound annual rate above 9 percent — driven by the same forces that made the Gulf states dependent on it: climate change, population growth, groundwater depletion, and the slow-motion realization that the planet’s freshwater supply was never distributed in a way that matches where humans decided to build cities.

    How it works and what it costs

    Two technologies dominate. Thermal desalination — primarily multi-stage flash distillation — heats seawater, evaporates it, and condenses the steam into freshwater. It’s the older method, consumes 5 to 12 kilowatt-hours per cubic meter, and remains prevalent in the Gulf because waste heat from co-located power plants can offset the energy cost. Reverse osmosis pushes seawater through semi-permeable membranes at high pressure, allowing water molecules through while blocking salt. RO consumes 2 to 4 kilowatt-hours per cubic meter — roughly half the energy of thermal methods — and has become the dominant technology globally because of that efficiency advantage.

    The cost trajectory has been dramatic. Twenty years ago, desalinated water cost roughly $1 per cubic meter. Over the last two decades, advances in membrane materials, energy recovery devices, and plant design have reduced that by approximately 80 percent. Recent bids in Abu Dhabi, Saudi Arabia, and Israel have come in below $0.50 per cubic meter for the first time. The Taweelah plant in the UAE — operational since 2022 with a capacity of 909,200 cubic meters per day — reportedly achieves costs as low as $0.49 per cubic meter. Israel’s Sorek II plant, producing 670,000 cubic meters daily, set a new record-low desalination water price when it was contracted. For context, the average American household uses roughly 1.1 cubic meters of water per day. At $0.50 per cubic meter, desalinated water costs the plant operator about 55 cents to produce a household’s daily supply. That’s not free, but it’s no longer prohibitive.

    The world’s largest desalination plant — Ras Al-Khair in Saudi Arabia, commissioned in 2014 — produces nearly 3 million cubic meters per day using a hybrid of thermal and RO technology, at a construction cost of approximately $7.2 billion. Saudi Arabia is planning to more than double its capacity: the Shuaiba 3 expansion (600,000 cubic meters per day, $821 million, powered partly by captive solar PV) entered commercial operation in 2025. The Rabigh 3 project adds another 600,000 cubic meters per day. NEOM — the $500 billion planned city — contracted a 500,000-cubic-meter-per-day RO facility with Veolia and Itochu, designed to run entirely on renewable energy and meet 30 percent of the city’s anticipated water demand.

    What’s advancing

    The next-generation improvements target the three constraints that limit current RO: energy consumption, membrane fouling, and brine disposal. Energy recovery devices now capture up to 70 percent of the hydraulic energy from the high-pressure brine stream that would otherwise be wasted, feeding it back into the system. Modern pressure exchangers have cut the net energy cost of RO plants significantly, pushing some facilities toward the thermodynamic minimum of roughly 1 kilowatt-hour per cubic meter.

    Membrane materials are where the research intensity is highest. Graphene oxide membranes — exploiting graphene’s two-dimensional nanochannels for faster water transport with higher salt rejection — have demonstrated permeability improvements over conventional polyamide membranes in laboratory settings. Aquaporin-based biomimetic membranes, which mimic the protein channels that biological cells use to transport water, represent an even more radical approach. Both remain pre-commercial at scale. The gap between laboratory performance and industrial deployment in desalination membranes is measured in years to decades, not months — each new material must demonstrate durability, fouling resistance, and consistent performance across millions of cubic meters before operators will trust it in a plant that supplies a city’s drinking water.

    Solar-powered desalination is the integration that could change the economics fundamentally, particularly in equatorial regions where solar irradiance is high and freshwater is scarce. Photovoltaic-powered RO systems have demonstrated specific energy consumption as low as 0.3 to 0.36 kilowatt-hours per cubic meter — an order of magnitude below conventional thermal methods. The NEOM plant is the highest-profile test of this approach at scale. Solar thermal desalination — using concentrated sunlight to directly evaporate seawater — is simpler and potentially cheaper for small-scale applications, but achieves lower throughput and is further from industrial deployment.

    What doesn’t work yet

    The brine problem is the constraint nobody has solved at scale. For every liter of freshwater a desalination plant produces, it generates roughly 1.5 liters of concentrated brine that is 1.5 to 2 times saltier than the intake seawater. The standard disposal method is pumping it back into the ocean through diffuser systems. The environmental impact is real: the Gulf’s waters are now estimated to be 25 percent saltier than typical seawater, in part because of decades of concentrated brine discharge from hundreds of desalination plants in a semi-enclosed body of water. Marine organisms in discharge zones show stress responses. The long-term ecological consequences of turning the Gulf into an increasingly hypersaline environment are poorly understood because nobody studied the baseline before the plants were built.

    The energy dependency is the strategic vulnerability. Desalination is energy-intensive regardless of technology. Countries that rely on desalination for most of their drinking water are converting an energy problem into a water problem — or, more precisely, coupling the two so that a disruption to energy supply becomes a disruption to water supply. The Gulf states’ pivot toward solar-powered desalination is partly an efficiency play and partly a hedging strategy: if oil revenues decline or fossil fuel supplies are disrupted, the water infrastructure needs an energy source that doesn’t depend on the same commodity the region exports.

    Forward osmosis, membrane distillation, electrodialysis, and various hybrid configurations are in active development — each targeting specific niches where conventional RO is suboptimal (brackish water, high-salinity environments, waste heat recovery). None has displaced RO as the dominant technology, and the pattern in desalination innovation is consistent: new approaches demonstrate promising laboratory results, face years of scale-up challenges, and either find a niche application or fail to compete with incrementally improving RO. The technology isn’t waiting for a breakthrough. It’s improving through accumulation — better membranes, better energy recovery, better pretreatment, better plant design — each shaving fractions of a kilowatt-hour or fractions of a cent off the cost per cubic meter.

    The honest constraint

    Desalination can produce unlimited freshwater from the ocean. That sentence is technically true and practically misleading. It can produce freshwater at a cost — in energy, in capital, in environmental impact, in operational complexity — that is falling but not zero, and that scales with the volume of water a society needs. Israel, which now gets roughly 80 percent of its domestic water from desalination, is the proof of concept: a technologically advanced country with high per-capita income that invested systematically in desalination infrastructure over 20 years and fundamentally solved its water scarcity problem. Whether that model is replicable in countries with lower per-capita income, weaker institutions, and higher water demand — India, Pakistan, sub-Saharan Africa — is the question that determines whether desalination solves the global water crisis or remains the solution for countries rich enough to afford it.

    We cover desalination alongside fusion energy, solid-state batteries, asteroid mining, and 20 other technologies racing to cross the gap between “works in principle” and “works on a Tuesday” across our Technology Moonshots course — where “done” means boring, measurable, and operable at scale, and desalination is the moonshot closest to actually being done.

  • Robot Dogs in 2026: From Boston Dynamics Spot to Battlefield Weapons

    In November 2024, the U.S. Secret Service deployed a Boston Dynamics Spot robot to patrol the grounds of Mar-a-Lago ahead of the president-elect’s arrival. No press release. No announcement. A photograph surfaced, circulated briefly, and the news cycle moved on. A few months later, in February 2025, Unitree’s founder and CEO Wang Xingxing shook hands with Xi Jinping at a meeting in Beijing, with Huawei’s CEO standing nearby. Both companies make four-legged robots. One costs $74,500 and signed a pledge never to weaponize its products. The other costs under $3,000 and has been showing up in PLA urban warfare exercises with rifles strapped to its back. The robot dog market in 2026 is not a single story about a single technology. It’s two stories diverging at speed — an American company trying to build a commercial inspection platform while refusing military applications, and a Chinese ecosystem that has already crossed the weapons line and is scaling production at a price point the West can’t match.

    The Spot economy

    Boston Dynamics’ Spot has been commercially available since 2020, initially at $74,500 per unit. The robot weighs 75 pounds, roughly the size of a German Shepherd, runs for approximately 90 minutes on a battery charge, climbs stairs, opens doors, and can be programmed for autonomous patrol routes. Its primary commercial use case has turned out to be industrial inspection — walking repetitive routes through data centers, oil refineries, construction sites, and utility infrastructure, capturing thermal data, detecting anomalies, and flagging maintenance issues without human fatigue or scheduling constraints.

    The data center market has become Spot’s growth engine. Boston Dynamics’ senior director of product management told Bloomberg in March 2026 that the company has seen a dramatic surge in data center interest, which makes sense: large, flat facilities with consistent patrol routes, equipment that benefits from continuous thermal monitoring, and 24/7 operational cycles where human fatigue creates real gaps. More than 60 bomb squads and SWAT teams across the U.S. and Canada now use Spot for hazmat incidents, armed standoffs, and hostage rescues — situations where sending a human or a real dog could be lethal. The NYPD briefly deployed a Spot unit (immediately nicknamed “Digidog”) in 2021, generating enough public backlash to force a cancellation, though the department later quietly reacquired the technology.

    In October 2022, Boston Dynamics signed an open letter pledging not to weaponize its robots or enable others to do so. Five other robotics firms co-signed. The pledge was voluntary, non-binding, and represented something unusual in defense technology: a company with a product the military obviously wants, choosing to draw a line. In February 2026, CEO Robert Playter retired after 30 years with the company, replaced by interim CEO Amanda McMaster. Whether the weapons pledge survives a leadership transition at a company owned by Hyundai — a conglomerate with its own defense interests — is an open question that nobody at Boston Dynamics has publicly addressed.

    The Chinese price point

    Unitree Robotics, founded in 2016 by Wang Xingxing, built its entire business model on being the affordable alternative to Boston Dynamics. Its Go2 consumer robot dog retails for under $2,000. Its B2 industrial model competes with Spot at a fraction of the cost. Wang has explicitly said Boston Dynamics is not his direct competitor — they took five years to release one product; Unitree releases one or two per year. The speed-and-cost advantage is real and the gap is widening.

    Unitree signed the same 2022 anti-weaponization pledge as Boston Dynamics. The company says it does not sell to China’s military. But a Kharon investigation in October 2025 found that Unitree has sold products to nearly 30 Chinese universities over five years, many of which have documented ties to PLA research programs and histories of providing equipment to military units. The PLA has conducted live training exercises featuring Unitree-style robot dogs advancing alongside infantry in urban warfare drills. Procurement records offer the paper trail; the company’s public denials offer the deniability. Wang Xingxing joined the advisory council to the U.S.-sanctioned chief executive of Hong Kong. The distance between “we don’t sell to the military” and “our products appear in military exercises through university intermediaries” is the distance the United Front Work Department has been navigating for decades — civilian-military fusion with plausible deniability built into the procurement chain.

    Chinese defense firms beyond Unitree — Deep Robotics, AeroArc, Xian Supersonic Aviation Technology — are building purpose-built military quadrupeds with rifles, grenade launchers, and autonomous targeting systems at unit costs below $30,000. The PLA isn’t waiting for the ethical debate to resolve. It’s fielding robot dog squads at a price point that makes mass deployment economically trivial. A Spot costs $74,500 and won’t carry a weapon. A Chinese military quadruped costs $30,000 and already has one mounted.

    Ghost Robotics and the American weapons question

    The American company that crossed the weapons line isn’t Boston Dynamics. It’s Ghost Robotics, a Philadelphia-based firm that has integrated rifles, sensors, and autonomy stacks onto its Vision 60 quadruped platform. Ghost Robotics has military contracts, has demonstrated armed configurations at defense trade shows, and has positioned itself as the company willing to do what Boston Dynamics won’t. The Vision 60 has been evaluated by the U.S. Air Force for base perimeter security and by the Department of Homeland Security for border patrol applications.

    Ghost Robotics represents the market reality that Boston Dynamics’ pledge can’t contain: if one company won’t weaponize its robots, another company will — and the customer (the Department of Defense) will buy from whoever says yes. The autonomous weapons debate that plays out in academic conferences and UN working groups plays out differently in defense procurement offices, where the question isn’t whether armed robot dogs are ethical but whether the adversary already has them.

    What 2026 looks like

    The installed base is growing across every sector simultaneously. Defense tech funding exceeded $28 billion in 2025 — up 200 percent year over year — and quadruped platforms are a visible beneficiary. ICE spent $78,000 on a robot for tactical operations. The German Bundeswehr demonstrated Spot at the Hannover Messe industrial trade fair. NATO-aligned countries are exploring robotic sentries for border monitoring. Japan and South Korea are testing robotic mobility assistants for confined military environments. India’s defense startups — Addverb Technologies, Svaya Robotics — are building quadruped platforms under the “Make in India” initiative.

    The price asymmetry between American and Chinese platforms is the strategic reality that matters most. Boston Dynamics builds a $74,500 inspection robot that won’t carry a weapon. Unitree builds a $2,000 consumer robot whose industrial variants cost a fraction of Spot and whose technology appears in PLA exercises. Chinese defense firms build purpose-built military quadrupeds for $30,000 with weapons already integrated. The unit economics enable deployment at a scale that overwhelms conventional defenses — which is the same cost-asymmetry logic that drives loitering munitions and drone swarms, applied to ground platforms.

    Boston Dynamics auditioned on Season 20 of America’s Got Talent in May 2025, performing a dance routine with Spot robots. The same month, PLA units were conducting urban warfare exercises with armed quadrupeds. The two images — dancing robots on a talent show stage and armed robots advancing alongside infantry — are the split-screen that defines the robot dog market in 2026. The technology is the same. The applications have already diverged. The question isn’t whether robot dogs will be weapons. It’s whether the company that builds the best one gets to decide.

    We cover robot dogs alongside the humanoid robot race, warehouse automation, healthcare robots, and the full spectrum of machines taking physical form across our Humanoid Robots & Drones course — where the question isn’t what robots can do but who decides what they’re allowed to.