Nuclear fusion has been “thirty years away” for roughly sixty years, which — if you’re keeping score — means the original timeline has now lapped itself twice. The joke is evergreen because the physics is genuinely hard: you need to confine a plasma at 150 million degrees Celsius long enough for hydrogen isotopes to fuse, extract more energy from the reaction than you put in to heat and confine it, do it reliably thousands of times in a row, connect it to a turbine, and keep the whole system running on a random Tuesday without a team of PhDs babysitting it. That last part — the Moonshot Tech course calls it “Tuesday-proof” — is where most fusion approaches are currently failing. The physics works. The demos work. The press releases definitely work. What doesn’t work yet is a machine that produces net electricity, sends it to a grid, and keeps doing that for 30 years while someone files maintenance reports.
But here’s what changed: $7.1 billion in private capital has flowed into fusion startups. Three companies have secured power purchase agreements with major tech firms. SPARC is being assembled. Helion broke ground on its first commercial facility. The National Ignition Facility achieved ignition. And the field has fractured into at least five fundamentally different approaches to the same problem — tokamaks, stellarators, field-reversed configurations, Z-pinch, inertial confinement — each with companies that believe their approach will get to “done” first. Here’s where each of them actually stands, ranked by proximity to a working power plant rather than proximity to a working pitch deck.
1. Commonwealth Fusion Systems — The Frontrunner
CFS is the closest thing the fusion industry has to a consensus leader, and the gap between first and second place is not small. The company, spun out of MIT in 2018, is building SPARC — a compact tokamak that uses high-temperature superconducting magnets made from REBCO (rare-earth barium copper oxide) tape to generate magnetic fields strong enough, according to the company, to lift an aircraft carrier out of the water. The magnets are the breakthrough that makes the economics plausible: stronger magnetic fields mean smaller reactors, which means lower construction costs, which means the path from “demo” to “product” is shorter than it was for previous generations of tokamak designs like ITER.
SPARC is currently under assembly at CFS’s headquarters in Devens, Massachusetts. The first of 18 D-shaped superconducting magnets has been installed. CFS plans to complete the magnet ring by summer 2026 and achieve first plasma in 2027. If SPARC demonstrates net energy gain from a privately built machine, it will be the most important milestone in fusion history — and the catalyst for CFS’s commercial plant, ARC, a 400-megawatt facility planned for construction near Richmond, Virginia, with a target online date in the early 2030s. Google has agreed to buy half of ARC’s output. Eni, the Italian energy company, has committed more than $1 billion. CFS has raised nearly $3 billion total, including an $863 million Series B2 round in August 2025 that was described as the last raise before SPARC demonstrates net energy. At CES 2026, CFS unveiled a digital twin of SPARC built in collaboration with Siemens and Nvidia. The company’s CEO, Bob Mumgaard, has said SPARC will be “nearly complete” by the end of 2026.
The risk: SPARC hasn’t produced plasma yet. The magnets are installed but the system hasn’t been tested as an integrated machine. The history of fusion is littered with devices that worked as components and failed as systems. CFS’s timeline — first plasma in 2027, commercial plant in the early 2030s — is aggressive by fusion standards and conservative by Elon Musk standards, which probably puts it in the right zone. But “nearly complete” and “producing net energy” are separated by an engineering chasm that has swallowed every previous tokamak program. The rare earth supply chain for REBCO tape is itself a constraint — the same critical materials bottleneck that our Rare Earth Elements course covers in detail applies directly to CFS’s magnet production pipeline.
2. Helion Energy — The Most Aggressive Timeline
Helion is the company most likely to be either spectacularly right or spectacularly wrong, and the timeline for finding out is short. The company uses a fundamentally different approach from CFS: a pulsed field-reversed configuration where magnets surround an hourglass-shaped chamber, plasma is spun into doughnut shapes at each end, the doughnuts are fired toward each other at more than a million miles per hour, and when they collide in the middle, the fusion reaction boosts the plasma’s own magnetic field, which induces an electrical current directly in the reactor’s magnetic coils. That last part is the key innovation — Helion’s design converts fusion energy directly into electricity without the intermediate step of heating water to drive a steam turbine. If it works, it eliminates the most expensive and maintenance-intensive component of a conventional power plant.
In July 2025, Helion broke ground on Orion — described as the world’s first commercial fusion power facility — in Malaga, Washington. Microsoft has agreed to purchase power from Orion, with a target delivery date of 2028. Sam Altman led the investment. The company’s seventh-generation prototype, Polaris, has demonstrated measurable deuterium-tritium fusion and achieved plasma temperatures of 150 million degrees Celsius. The Omega manufacturing facility, which will produce the thousands of capacitor units required for Orion, is expected to begin production in 2026.
The risk: Helion’s 2028 target for delivering power to Microsoft is the most aggressive commercial timeline in the industry by a significant margin. The company has demonstrated fusion at laboratory scale but has not demonstrated net electricity production. The gap between “measurable fusion” and “net electricity delivered to a data center” is the gap between a campfire and a power plant. Helion’s direct-conversion architecture is genuinely novel, which means it doesn’t have the decades of experimental validation that tokamak designs benefit from. If the approach works, it’s a paradigm shift. If it doesn’t, Microsoft’s data centers will need power from somewhere else. The operational question — whether a pulsed system firing plasma doughnuts at each other millions of times can maintain reliability over years of continuous operation — is completely unanswered. That’s the Battlefields of the Future problem applied to energy: the demo works great, the question is whether it survives contact with reality at scale.
3. TAE Technologies — The Oldest Startup, Now Merging with Trump Media
TAE Technologies — founded in 1998 at UC Irvine, making it the oldest private fusion company in existence — pursues what may be the most scientifically ambitious approach in the field: hydrogen-boron fusion. Most fusion companies use deuterium-tritium fuel, which fuses at relatively accessible temperatures but produces high-energy neutrons that damage reactor walls, activate structural materials, and create radioactive waste that requires shielding and long-term management. Hydrogen-boron fuel produces no neutrons. If TAE can make it work, the reactor engineering becomes dramatically simpler: no neutron shielding, minimal radioactive waste, and a reactor that is fundamentally safer to operate and decommission. The catch is that hydrogen-boron fusion requires temperatures roughly ten times higher than deuterium-tritium — approximately 1 billion degrees Celsius — which is why no one else is trying it.
TAE’s current machine, Norman (named after founder Norman Rostoker), uses a field-reversed configuration where two plasma shots collide and are then bombarded with particle beams to maintain stability. Google has been a technology partner for more than a decade, contributing machine-learning algorithms for plasma control. The company has raised approximately $1.79 billion. Its next-generation machine, Copernicus, is designed to reach the temperatures required for hydrogen-boron fusion. The commercial plant, Da Vinci, targets grid-ready electricity in the early 2030s.
The wildcard: in December 2025, Trump Media & Technology Group — the company behind Truth Social — announced an all-stock merger with TAE Technologies valued at more than $6 billion. TAE plans to begin building a 50-megawatt utility-scale fusion plant in 2026, aiming to generate electricity by 2031. The merger, if it closes as planned in mid-2026, would make TAE one of the world’s first publicly traded fusion companies. The political valence of a fusion company merging with a media company controlled by a sitting president is — to use a term from our Shadowcraft course — the kind of institutional entanglement that deserves its own lecture. Whether the merger accelerates TAE’s engineering timeline or distracts from it is the open question. The physics doesn’t care who owns the stock.
4. Zap Energy — The Simplest Machine
Zap Energy’s approach is appealingly minimalist: no superconducting magnets, no cryogenics, no high-powered lasers. The company uses a sheared-flow-stabilized Z-pinch — an electromagnetic phenomenon where electric currents sent through plasma generate magnetic fields powerful enough to compress the plasma to fusion conditions. The plasma essentially confines itself. The reactor is compact, the component list is short, and the cost per experimental cycle is dramatically lower than any tokamak or stellarator program. In November 2025, Zap demonstrated pressure levels 10,000 times atmospheric pressure at sea level, a key milestone in validating the Z-pinch approach.
Zap has raised $337 million from Bill Gates’ Breakthrough Energy Ventures, Chevron Technology Ventures, and others. The company is based in Everett, Washington — the same city as Helion, making Everett arguably the world capital of non-tokamak fusion. Zap’s advantage is iteration speed: cheaper experiments mean faster learning cycles. The disadvantage is that Z-pinch confinement is inherently less stable than tokamak or stellarator confinement, and scaling the approach from laboratory demonstration to power-plant operation introduces engineering challenges that the simplicity of the concept doesn’t eliminate. The machine is simple. The physics of keeping a self-compressing plasma column stable at power-plant scales is not.
5. The Stellarator Renaissance
Stellarators are the dark horse of the fusion race, and 2025 may have been the year they became the smart money’s second bet. Germany’s Wendelstein 7-X — the world’s most advanced stellarator, operated by the Max Planck Institute — achieved record energy turnover and sustained high-performance plasma operation in 2025, demonstrating that stellarators can maintain the long-duration stability required for baseload power generation. Unlike tokamaks, which require a pulsed electrical current to maintain plasma stability (creating engineering complications for continuous operation), stellarators use externally generated magnetic fields that can theoretically run indefinitely without pulsing. The tradeoff has always been complexity: stellarator geometries are twisted, asymmetric, and extraordinarily difficult to design, manufacture, and assemble. What changed is computation — AI and modern simulation tools now allow engineers to optimize stellarator geometries that would have been impossible to design even a decade ago.
Three private stellarator companies have emerged as credible contenders. Proxima Fusion, based in Munich, raised €130 million and unveiled its Stellaris power-plant architecture. Type One Energy, which has raised $269 million, is planning to build a 350-megawatt stellarator on the site of a retired TVA coal plant in Tennessee — a location choice that says something about where fusion fits in the energy transition. Thea Energy completed the first early design review in the DOE’s Milestone-Based Fusion Development Program, using roughly 350 planar superconducting coils instead of a few complex magnets — a manufacturing simplification that could make stellarators buildable at industrial scale.
Stellarators are further from commercial deployment than CFS or Helion — most target the early 2030s for prototype operation — but they may be the architecture best suited for what the grid actually needs: continuous baseload power that runs for decades without pulsing. The humanoid robotics industry faces a version of the same challenge — the demo that works for 90 seconds on stage and the product that works for 90,000 hours in a warehouse are different engineering problems. Stellarators are betting on the warehouse version.
6. Everyone Else Worth Watching
Pacific Fusion raised a $900 million Series A in 2025 — one of the largest first rounds in fusion history — for an inertial confinement approach using electromagnetic pulses rather than lasers. General Fusion, the Canadian company that nearly ran out of money in spring 2025, survived through emergency funding rounds and is now going public via a reverse merger that could bring in $335 million. Tokamak Energy, a UK-based company pursuing a spherical tokamak design, raised $125 million and maintains credible engineering without setting aggressive commercial dates. SHINE Technologies has taken the most pragmatic approach in the sector: rather than waiting for power-plant-scale fusion, SHINE generates revenue today from fusion-adjacent applications including medical isotopes and industrial inspection, making it one of the only fusion companies with actual cash flow. Xcimer Energy is building laser systems five times more powerful than the National Ignition Facility’s equipment, targeting the repetition-rate problem that separates NIF’s single-shot ignition achievement from continuous power generation.
The honest scorecard
Here’s where the field actually stands, stripped of press-release language:
Has any private company demonstrated net energy gain from fusion? No. CFS is closest. SPARC’s first plasma is targeted for 2027. If it works, that changes everything. If it doesn’t work on the first attempt, CFS likely iterates — the company has the capital and the engineering depth to troubleshoot.
Has any private company produced net electricity from fusion? No. Helion claims it will by 2028. The consensus outside Helion’s investor presentations is early 2030s at the most optimistic. A decade ago, the consensus was “never, probably.” The consensus has moved.
Is fusion closer to reality than it was five years ago? Unambiguously yes. The NIF ignition result in December 2022 settled the physics question — fusion can produce more energy than it consumes. The engineering question — can you do it reliably, affordably, and at scale? — is the question the $7.1 billion is trying to answer. The neuroprosthetics field went through a similar inflection: the science was proven decades ago, but the engineering required to turn laboratory BCIs into devices people use 10 hours a day took another twenty years of iteration. Fusion may be on a similar curve — the physics is settled, the engineering is the bottleneck, and the capital is finally flowing at the scale the engineering requires.
What could kill the timeline? Supply chain constraints on high-temperature superconducting tape, tritium availability (the global supply is approximately 25 kilograms, nearly all of it produced as a byproduct of Canadian CANDU reactors), regulatory frameworks that don’t yet exist for commercial fusion plants, and the possibility that plasma instabilities at power-plant scale behave differently than plasma instabilities at experimental scale. Also, the general principle that any technology whose advocates have to repeatedly insist “this time it’s different” should be evaluated with the same skepticism we’d apply to any other extraordinary claim.
What would make a fusion skeptic change their mind? SPARC achieving net energy. Helion delivering electricity to Microsoft. Any company demonstrating sustained, repeatable net electricity production outside a government laboratory. Those milestones are scheduled for 2027-2028. We’ll know within two years whether the thirty-year joke needs updating — or whether it gets another thirty years of shelf life.
This is the kind of technology our Moonshot Tech course was built to evaluate — where a field that has been promising the same thing since the 1960s is suddenly backed by $7.1 billion in private capital, three power purchase agreements with the world’s largest tech companies, and a machine in Massachusetts whose magnets are theoretically strong enough to lift an aircraft carrier, all converging on a two-year window that will either vindicate the optimists or add another chapter to the longest-running joke in energy.