The hinge of the story of nuclear fusion is energy need. In rich and poor countries alike, electricity has become the bloodstream of modern life. Cars are slowly becoming batteries on wheels. Heat pumps and induction stoves replace burners. Factories turn to electric furnaces and electrolysers. And then there is the demanding new citizen of the grid: the data centre, swollen by artificial intelligence, hungry for steady power at all hours. Analysts at international agencies now talk about ‘firm clean power’ the way city planners once spoke of water mains—something so dependable that the rest of the system can relax around it. Wind and solar, abundant and cheap, will shoulder much of the load. But even their fiercest advocates concede that the world will need dense, weather‑independent sources to backstop them. Fusion’s pitch is not to replace renewables but to join them, quietly and continuously, with fuel extracted from seawater and from the light metal lithium.
If fusion is such a tidy idea on paper, why did it take so long to stand up in the lab? Because the Sun cheats. A star uses gravity itself to press atoms together. On Earth, engineers are forced to counterfeit gravity with machines. One family of machines builds a kind of magnetic bottle—a force field sculpted so that a 100‑million‑degree gas never touches the walls. Nothing physical can hold such heat; only the magnetism does. The other family gives up on long duration and chooses violence. It takes a tiny frozen bead of fuel and, with a bank of lasers, smashes it from all sides so that for the blink of a blink the fuel is denser than lead and hot enough to ignite. In one approach you keep the fire going, like a chef nursing coals; in the other you create it in pulses so brief they must be counted in nanoseconds. Both ideas are older than most of their current practitioners. The new thing is not that they work once—it is that they work repeatedly, with the quality control of a manufacturing line.
ITER: the world’s big bet on “proof at scale”
On a scrubbed patch of Provence, 35 governments are assembling ITER, a tokamak built to answer a single question at power-plant scale: can a magnetically confined plasma heat itself enough to sustain a burn? ITER isn’t designed to produce electricity; it’s a physics and engineering demonstrator that aims to make ten times more fusion power than the external heating it absorbs—roughly 500 megawatts of fusion heat from about 50 megawatts injected into the plasma, for several-hundred-second pulses (the famous “Q≥10”). The project’s leadership adopted a revised “Baseline 2024” that sequences the machine toward substantive research operations in the 2030s (hydrogen and deuterium campaigns first, then full-bore phases), reflecting component and assembly realities after years of construction and supply-chain lessons. The near-term deliverable is not electrons to the grid, but a burning plasma and a library of solutions—divertor heat handling, wall materials, tritium accountancy—that industry can adopt
How Fusion Becomes an Industry — with Russia’s Nuclear Cities Back in the Frame
On certain mornings in California, the laser bays at the National Ignition Facility hum like an orchestra tuning up—thousands of parts aligning toward a single downbeat. In a control room behind thick glass, engineers speak softly into headsets and watch the last decimal places settle on their screens. When the countdown reaches zero and two megajoules of light crash into a target the size of a peppercorn, the room scarcely moves. There is no flash to see, only numbers to parse: the arc of a pressure wave, the neat rise of a neutron count, the stubborn, pleasing asymmetry of a pulse shaped exactly as intended. It is a quiet theater for a loud ambition—to bottle the process that lights the Sun.
That ambition has a surprising texture in 2025. For decades, nuclear fusion lived in the half‑imaginary realm reserved for flying cars and photographs of Mars bases—always arriving, never quite here. But across the world’s laboratories and factory floors, th tempo has changed from imagination and science fiction to routine.
Shots at the American laser facility that once made headlines now repeat with small, deliberate improvements. In Germany, a sinuous magnetic device holds plasmas steady for long stretches, its components actively cooled like a race car’s brakes. In Korea and China, doughnut‑shaped machines push toward the calm, long‑lasting operation that a power plant would need. Japan’s new superconducting machine is no longer a construction site but a living, humming instrument. The rhythm is not triumphant; it is industrious.
To understand what this matters for, it helps to strip the story down to its simplest picture. There are two kinds of nuclear power worth distinguishing. Fission, the kind we use today, is a willed unravelling of heavy atoms like uranium. The atom splits, heat is drawn off, and steam turns a turbine. Fission works, decarbonises grids and keeps lights on through windless nights, but it requires heavy safety systems and leaves wastes that must be shepherded for a long time. Fusion, by contrast, is a marriage of light atoms. Two forms of hydrogen—deuterium and tritium—are pressed together until they become one, releasing a fast neutron and a puff of harmless helium. There is no chain reaction to run away. If the super‑hot gas in which the reaction lives slips out of its narrow sweet spot, the fire goes out by itself. Radioactivity still enters the picture—materials near the action are peppered by neutrons, and tritium must be carefully accounted for—but the risk profile is different from the chemistry of fission’s long‑lived products.
Progress demanded that several hard things be done at once. The physics problem is to keep a tiny star hot, dense and calm long enough to matter—three knobs that refuse to be turned together. Raise the heat and the gas becomes twitchy; add density and the edge starts to spit; lengthen the run and the machine’s exhaust begins to glow with unwelcome intensity. The materials problem is more elemental. The parts near the plasma face are punished by heat worthy of a rocket nozzle and by an invisible hail of fast neutrons that rattle apart the tidy lattices of metals and ceramics. So researchers invent new alloys and new ways to carry heat away: channels bored through steel like the veins of a leaf; tiles that shrug off thermal shocks; liquid‑metal blankets that catch neutrons, breed tritium and protect what lies beneath. And then there is the control problem. A high‑performance plasma wriggles like a living thing. To keep it well‑behaved, operators nudge it constantly with radio waves, with microwaves, with little pellets of fuel and with currents in coils that tweak the shape of the magnetic field—a ballet increasingly choreographed by software. For years, laboratories around the world could produce heroic single shots. What they could not do was hold that quality steady and high at the densities a power plant must inhabit. That is the change of the past two years: the heroics have learned to repeat.
The road to this point is littered with admirable detours. American ‘mirror’ machines once seemed the future; they gave way to magnetic doughnuts when the physics turned against them. Stellarators—the doughnut’s baroque cousin—were set aside in many countries before their careful rebirth. Laser fusion chased the idea of ignition for longer than some of its champions care to remember, and then, suddenly, found itself there more than once. None of it was wasted. Spitzer’s early sketches, Lawson’s famous criterion, the transport scalings that sound like poetry to the uninitiated—they all became parts of the modern toolkit.
The history is not only scientific; it is political and cultural.
Nuclear power rose quickly in the 1960s and 70s—helped along by oil shocks—and then lost public sympathy after the accidents at Three Mile Island, Chernobyl and Fukushima. Even as fleets of existing reactors delivered quiet decarbonisation, new projects faltered on cost and fear. In the past few years, that posture has softened. Climate targets have deadlines. Gas pipelines can be turned off. Air conditioners proliferate in hot cities. Policymakers began to speak again about nuclear power—first about life‑extension for old plants, then about new small designs, and finally about fusion, which benefits from nuclear’s professional backbone while promising a different kind of risk.
Meanwhile the machinery grew up. The American laser facility now has a ledger of ignition shots rather than a single trophy; the numbers wobble less, the design tools predict more. In Germany, a stellarator that looks like sculpture has been standing up long, placid plasmas while water‑cooled innards carry off the heat. In Korea, a tungsten‑armored machine holds high‑performance operation for over a minute without the damaging edge bursts that once ended runs with a nasty crackle. In China, a laboratory famous for persistence reports thousand‑second operation in improved regimes and, in fresh work, a way to push to higher densities without losing grace. Japan’s big superconducting device, built with Europe, is trading hard hats for lab notebooks. Elsewhere, private companies are doing the unglamorous things that mark the beginning of an industry: ordering transformers, pouring concrete, buying insurance, signing power‑purchase agreements. Investors have become choosier than they were a few years ago, but governments in America and Britain have answered with strategies and planning rules that treat fusion as something to be sited and permitted, not merely admired.
The Closed Secret Cities of Russia
One chapter in this story begins, as so many twentieth‑century chapters did, in a closed city. In the late 1960s, results from a Soviet machine in Moscow were so striking that British physicists flew in with their own instruments to check them. The measurements stood up, and in a season the world changed course. The doughnut‑shaped ‘tokamak’ became the ship of choice for magnetic fusion. For a while after the Soviet Union’s collapse, the country’s once‑integrated nuclear archipelago—an archipelago of secret towns tucked into forests, each with a bland official name and a sharper nickname—seemed to recede. But in the past few years Russia’s programme has taken on a particular character again, one that fits the times. It is less interested in scoring records than in building the guts of a plant. A rebuilt research machine in Moscow has been used to practise the day‑to‑day chores of operation: heating, fueling, keeping the plasma tidy over longer periods. A next‑step design aims not at glory but at endurance, a long‑running testbed to punish exhaust systems and fuel‑cycle equipment under realistic heat and neutron loads. In Troitsk, a town south of the capital, engineers are standing up facilities to do precisely that kind of punishment testing. In St. Petersburg and elsewhere, companies produce the high‑power microwave sources—gyrotrons—that heat plasmas not just at home but in international projects. The ecosystem that once wound coils for the first tokamak has remembered how to make parts other people need.
Those parts are reminders that fusion, like aviation or chipmaking, is a supply‑chain sport. Beyond the physics, there are vendors of superconducting tapes and cryogenic valves, makers of ultra‑high‑vacuum vessels, specialists in power supplies that can sling thousands of amps with a surgeon’s poise. A generation of engineers trained on Europe’s great experimental device has dispersed into industry programmes, while in India companies honed the gentle violence required to build a vacuum vessel the height of a cathedral. The United States and Britain, for their part, have produced the political paperwork that industries quietly crave: a national strategy here, a planning statement there, the dry alchemy by which uncertainty becomes momentum.
Is fusion cheap? Not yet, and that is the honest answer that people in the field now give without flinching. The fuel is cheap—deuterium can be distilled from seawater, tritium can be bred from lithium inside the plant—and there is no carbon in the exhaust. But the machines themselves are demanding and, at first, expensive. Costs will depend on how long their innards last: the superconducting magnets, the microwave sources, the components that breathe in heat and exhale it into cooling loops. Laser‑driven schemes must master the art of inexpensive, perfect little targets made by the million. Economists who build models for a living say the technology will have to fall into the same cost band as modern nuclear and as packages of renewables plus storage if it wants to compete in ordinary markets. That has been true of every new energy technology from combined‑cycle gas to offshore wind; the first copies are dear and the tenth are not. On sustainability, the shape of the task is also clear: keep a tight ledger on tritium, choose materials that are easy to recycle, and design components so that what radioactivity they do pick up wanes to low hazard within decades rather than millennia.
What comes next depends less on miracles than on persistence. The largest leaps in science and engineering often arrive in what officials like to call ‘mission mode’. The wartime project that industrialised fission remains the grim archetype; the vast accelerators under fields of European wheat and the rockets that touched the Moon are its more cheerful heirs. Today’s fusion programmes echo that posture. Chinese laboratories, European consortia, Japanese and Korean teams and Russian institutes have all chosen lanes and begun to run them with a runner’s unromantic discipline: build hardware; operate; fix; repeat. The likeliest near‑term future is incremental and, for that very reason, believable. Several pilot‑class machines will reach their first plasmas in the late 2020s. One or two will, in the early 2030s, show electricity to the grid for hours or days at a time. By the latter half of the decade, a small number of sites may sell firm, clean power under contracts to data‑centre campuses or clusters of industry. It will not feel like a moon landing. It will feel like a craft being learned in public.
And that, finally, is why the quiet theater of the laser bay in California, the spare beauty of a German stellarator’s curves, the workaday heat‑exchanger designs sketched in Russian offices, all belong to the same story. For years fusion was treated as a riddle with a single dramatic answer. It is turning out to be a trade—a thousand answers stacked neatly in parts bins, a few dozen in software repositories, some etched into muscle memory by technicians who no longer think of themselves as pioneers. What lights the Sun is not a mystery. What remains is to build the habit of tending that fire. Crafts, once learned, are hard to forget.
Who’s closest to “steel in the ground” for commercial fusion
Commonwealth Fusion Systems (CFS) — The MIT spin-out that vaulted HTS magnets from lab to factory floor now says the compact SPARC experiment is entering assembly/commissioning as the physics proof for its first plant, ARC. Virginia has been selected for the 400-MW grid plant, targeting the early 2030s, and CFS has begun the unglamorous but crucial work: site selection, interconnect planning, and long-lead procurement. Recent reporting notes multi-hundred-megawatt commercial PPAs and an offtake path that treats fusion as firm power for a data-center corridor. The short version: magnets first, then a first-of-a-kind plant with a conventional steam cycle.
Helion Energy — Helion’s pulsed, magneto-inertial approach compresses and heats plasma in rapid bursts and converts energy directly to electricity. It signed the world’s first fusion power-purchase agreement (with Microsoft) and has started site work in Washington state; local authorities have now green-lit the next phase of construction. Helion’s narrative is unusually commercial—grid tie-in near existing infrastructure, a marketer (Constellation) to handle delivery, and a named customer with a date certain. The technical hinge is migrating what they learn on the “Polaris” prototype to the first grid machine (“Orion”) on a tight schedule.
TAE Technologies — The California veteran of “beam-driven” field-reversed configurations is building Copernicus, the sixth in its machine series, with the stated goal of demonstrating net-energy viability before moving to a power-plant design. TAE’s recent updates emphasize simpler startup and sustained high-temperature plasmas, alongside fresh financing to carry Copernicus through commissioning. If they can hold temperature and stability while tightening the energy balance, their follow-on (often dubbed “Da Vinci”) is framed as an engineering scale-up, not a physics gamble.
General Fusion — Pursuing “magnetized-target” fusion with a liquid-metal liner, General Fusion pivoted from a Culham visitor-center demo to its LM26 program and is now touting a stepwise path: first plasma achieved on a new machine, peer-reviewed modeling of confinement time, temperatures past the 100-million-degree mark, and a breakeven target staged for the mid-2020s, with commercialization in the 2030s. The appeal here is a potentially forgiving liquid wall and a plant layout that looks familiar to heavy industry. The burden is proving repetition and efficiency.
Tokamak Energy — A UK pioneer of spherical tokamaks and HTS magnets, Tokamak Energy positions ST80-HTS (a high-field, compact machine) as the technical bridge to ST-E1, a pilot plant designed to deliver tens of megawatts of net electricity in the early 2030s. Their recent concept notes point to 800 MW of fusion power translating to ~85 MW net electric in a first-plant configuration—conservative on net power, but clear on learn-by-doing. The company’s strategic edge is magnet IP and a design that aims to be manufacturable.
Zap Energy — Eschewing big magnets, Zap pursues a sheared-flow Z-pinch in a compact device line (FuZE, FuZE-Q). The story here is cadence: tens of thousands of plasmas a year, rising neutron yields, and a system built for rapid iteration. It’s earlier on the commercialization timeline than CFS/Helion, but if the team can stabilize at higher parameters without bulky magnets, the plant footprint could be very small.
Type One Energy — A stellarator startup born of veteran W7-X talent, Type One is refurbishing a Tennessee Valley Authority site for Infinity One, an in-situ stellarator testbed with TVA and Oak Ridge as neighbors. The goal is to de-risk manufacturability, maintenance and uptime—the prosaic virtues a utility demands—on the way to a pilot. Stellarators trade startup pulses for continuous operation, making them catnip for grid planners—if you can build them repeatably.
Proxima Fusion & Renaissance Fusion (EU stellarator wave) — Munich-based Proxima, a Max-Planck spin-out, has raised Europe’s biggest fusion round this year to pursue a stellarator plant, publishing early-stage open plans and targeting a model-coil by 2027 and a demonstration in the early 2030s. Grenoble-based Renaissance is betting on direct-deposited HTS coils and flowing liquid-metal walls to simplify stellarators and harden them for plant service. The European bet: build on Wendelstein-7X know-how and turn “always-on” into a product spec.