Can We Ignite a Star on Earth as a Practical Power Source?

On the morning of December 5, 2022, inside a target chamber at the National Ignition Facility in Livermore, California, 192 laser beams converged on a capsule of frozen hydrogen isotopes smaller than a peppercorn. For a few trillionths of a second, conditions inside that capsule exceeded the density and rivaled the temperature at the center of the sun. Out of 2.05 megajoules of laser energy poured in, 3.15 megajoules of fusion energy came out. It was a genuine milestone — the first time a fusion experiment produced more energy than the light that triggered it — and headlines around the world declared, not for the first time in fusion's history, that limitless clean energy was finally at hand.

It wasn't, quite. The 2.05 megajoules counted only the energy in the laser light itself; the lasers had drawn over 300 megajoules from the wall plug to produce that light, thanks to notoriously inefficient flashlamp pumping. The plant, in other words, still spent roughly a hundred times more energy than it got back, once you count the whole system rather than just the spark. This is the strange, humbling truth at the center of fusion energy in the 2020s: humanity has now, unambiguously, ignited a self-sustaining thermonuclear burn on Earth. We have not come close to making one that pays its way.

That gap — between a laboratory flash that proves fusion can release net energy from the fuel, and a power plant that can do it every few seconds, for decades, cheaper than the alternatives — is one of the great open engineering-physics problems of our age. It is not a question of whether the nuclear physics works. It obviously does; it has been working in the sun for four and a half billion years. It is a question of whether we can tame a hundred-million-degree gas long enough, cheaply enough, and continuously enough to make a business out of starlight.

svg[The Lawson criterion: two strategies for the same triple product]

::: Both magnetic and inertial approaches aim to cross the same ignition threshold, trading confinement time against density.

By these measures, real progress has been made, and it deserves to be stated plainly rather than buried under caveats. The Joint European Torus (JET) in England achieved 69 megajoules of fusion energy over a five-second pulse in 1997 and, in its final experiments in late 2023 before decommissioning, 69 MJ again with an improved power profile, using a genuine deuterium-tritium fuel mix. NIF's 2022 shot, repeated and improved through 2023 and 2024, reached fusion energy gain (output over laser input) exceeding 1.5, then higher still. China's EAST tokamak has held a plasma above 100 million kelvin for over a thousand seconds. These are not incremental curiosities; they represent decades of hard-won progress in plasma control, diagnostics, and materials, following a trajectory that, measured in triple product, has actually outpaced Moore's Law over multi-decade stretches. The physics of igniting a fusion plasma is no longer in serious doubt. What remains genuinely uncertain is everything required to turn ignition into a power plant.

The trouble starts with the very quantity that makes fusion reactions attractive: the 17.6 MeV released per deuterium-tritium reaction is split, by the kinematics of the reaction, as roughly 3.5 MeV to the helium nucleus and 14.1 MeV to a free neutron. That neutron, uncharged and therefore immune to any magnetic bottle, flies out of the plasma and slams into the reactor's inner wall at nearly a fifth of the speed of light. This is, from an energy-harvesting perspective, wonderful — it is how the reactor's heat gets out to a coolant loop at all. It is also, from a materials-engineering perspective, close to a nightmare. Over the projected multi-decade life of a fusion power plant, the first-wall and blanket structures would absorb a neutron fluence far exceeding anything in a conventional fission reactor, whose neutron spectrum is softer and whose damage mechanisms are comparatively well characterized after seventy years of operational experience. Fusion neutrons, at 14.1 MeV, knock atoms clean out of their lattice positions, transmute stable isotopes into helium and hydrogen inside the metal, embrittle steel, swell components, and degrade the very properties — strength, ductility, thermal conductivity — the plant depends on for its lifetime. No material has ever been subjected to the neutron fluence and flux a commercial fusion first wall would need to survive for, say, thirty years, because no facility exists that can generate a high-flux 14 MeV neutron source for materials testing at industrial scale. This is not a hypothetical problem awaiting future funding; it is an acknowledged, unresolved gap in the materials science itself. Proposed facilities like IFMIF-DONES (International Fusion Materials Irradiation Facility) are meant to close it, but as of the mid-2020s they remain under construction or in planning, not yet delivering data.

There is a second, related problem: tritium. Deuterium is abundant and cheap, but tritium has a half-life of about 12.3 years and does not occur naturally in useful quantities — the world's entire inventory, mostly a byproduct of Canadian CANDU fission reactors, amounts to only a few tens of kilograms, and it is shrinking as those reactors retire. A commercial fusion plant is expected to breed its own tritium in situ, via neutron capture on lithium in a surrounding "blanket," but achieving a tritium breeding ratio reliably above 1.0 — producing at least as much tritium as the plant consumes, with margin for radioactive decay and processing losses — has never been demonstrated in an actual burning plasma. It is a problem of neutron economy, chemistry, and engineering combined, and it remains, in the candid assessment of most fusion scientists, unsolved in practice rather than merely unbuilt.

Then there is the plasma itself, which even after seventy years of study remains only partially tamed. Tokamak plasmas are susceptible to disruptions: sudden, catastrophic losses of confinement in which tens of megajoules can be dumped onto the vessel wall in a few milliseconds, sometimes accompanied by "runaway electrons" accelerated to near light speed by the collapsing electric fields, capable of punching through structural components. ITER, the international collaboration under construction in southern France with contributions from 35 countries, is explicitly designed around the expectation that disruptions will occur and must be mitigated, not eliminated — a sobering admission from the largest fusion experiment ever attempted. Edge-localized modes, instabilities at the plasma boundary that periodically eject bursts of heat and particles onto the divertor — the component designed to exhaust waste heat and helium ash — pose a related challenge: even in steady operation, the divertor of a reactor-scale tokamak must absorb heat fluxes comparable to those on a spacecraft reentering the atmosphere, but must do so continuously, shot after shot, for years.

Finally, the economic question, which is not really a physics question at all but may be the hardest of the four. Even a fusion plant that solves ignition, materials, and tritium breeding still needs to generate electricity more cheaply than its competitors. Fusion facilities are complex, capital-intensive, and — in the tokamak case especially — enormous; ITER's cost, spread across decades and international partners, is estimated in the tens of billions of dollars, and it is explicitly a research device, never intended to produce electricity at all. Whether a smaller, cheaper, mass-manufacturable fusion plant is even physically possible — whether the physics allows shrinking the machine while holding the triple product fixed — is itself a live design question, at the heart of the recent wave of high-field, high-temperature-superconductor-based compact tokamak proposals.

The landscape of serious fusion approaches in the mid-2020s can be organized less by nationality or funding source than by which piece of the Lawson criterion, and which engineering compromise, each approach is betting on.

The mainline tokamak, scaled up. ITER represents the culmination of seventy years of tokamak physics: a machine large enough (a plasma volume of 840 cubic meters, a major radius of 6.2 meters) that its energy confinement time should be long enough, at achievable density and temperature, to satisfy Lawson's criterion by a comfortable margin and sustain a burning plasma with a fusion gain Q of 10 — ten units of fusion power for every unit of external heating power. Its appeal is that it rests on the most thoroughly validated confinement physics available; its size follows from empirical scaling laws (the "IPB98" confinement scaling, derived by fitting data across dozens of tokamaks worldwide) that, while imperfect and debated, have proven reasonably predictive. Its trouble is that bigness is expensive, slow to build, and does not obviously map onto commercial deployment: ITER, first conceived in 1985, will not achieve full deuterium-tritium operation until the 2030s at the earliest, following repeated schedule delays, and it is a single experimental reactor, not a template for mass-produced power plants.

Compact, high-field tokamaks. A different bet, pursued by companies like Commonwealth Fusion Systems (spun out of MIT) with its SPARC device, and by the UK's STEP program, exploits a discovery in an adjacent frontier of physics: high-temperature superconductors, particularly rare-earth barium copper oxide (REBCO) tapes, which can sustain much higher magnetic fields than the low-temperature niobium-based superconductors used in ITER. Because fusion power density scales steeply with magnetic field strength — roughly as the fourth power in some simplified scaling estimates — doubling the field can shrink the machine dramatically while preserving the physics. SPARC, using a field of roughly 12 tesla at its magnet coils compared to ITER's 5.3 tesla, aims for net energy gain in a device many times smaller and, its designers hope, far cheaper and faster to build. This program is a direct beneficiary of progress in the neighboring frontier of high-temperature and exotic superconductors: without the industrial maturation of REBCO tape manufacturing over the past two decades, the compact-tokamak strategy would not exist. Its appeal is speed and cost; its risk is that confinement scaling at these more extreme parameters, and the behavior of disruptions and divertor heat loads in a smaller, more tightly packed device, are less thoroughly tested than in ITER's more conservative design.

The stellarator revival. For decades, stellarators were the road not taken — theoretically attractive for steady-state operation free of disruption risk, but so complex to design that early machines badly underperformed their tokamak cousins. That changed with the German Wendelstein 7-X, which began operation in 2015 after a construction and design process spanning nearly two decades, using supercomputer optimization to shape 50 non-planar superconducting coils into a form that minimizes the particle-transport losses that plagued earlier stellarators. W7-X has since demonstrated confinement quality approaching that of comparable tokamaks, alongside inherent steady-state stability and freedom from disruptions — arguably the single most dangerous failure mode of the tokamak line. Its trouble is manufacturing complexity: the coils must be built and aligned to sub-millimeter tolerances in three dimensions, a far harder engineering task than a tokamak's more symmetric geometry, and no stellarator has yet approached ignition-relevant plasma performance at reactor scale.

Inertial fusion energy. NIF's 2022–2024 results proved that laser-driven inertial confinement can produce scientific energy gain, but NIF was never designed as a power-plant prototype — its flashlamp-pumped lasers fire at most a few times per day, whereas a power plant would need repetition rates of roughly ten shots per second, using far more efficient diode-pumped lasers, along with a practical scheme for mass-manufacturing and injecting fuel pellets, and a chamber design that survives repeated micro-explosions indefinitely. Private companies including Focused Energy and Xcimer Energy, along with government-funded programs, are pursuing this path, betting that laser and target-fabrication technology can mature fast enough to close that gap. Its appeal is the demonstrated physics milestone and a conceptually simpler magnet-free confinement scheme; its trouble is that almost every other subsystem — laser efficiency, repetition rate, target manufacturing at industrial scale, and chamber survivability — remains far less mature than the equivalent tokamak subsystems.

Alternative and unconventional schemes. A wider ecosystem of private ventures is betting on architectures outside the tokamak-stellarator-laser trichotomy: field-reversed configurations (TAE Technologies), magnetized target fusion that compresses a magnetized plasma with pistons (General Fusion), and Z-pinch approaches (Zap Energy) that use the plasma's own current to self-compress without external magnets at all. These programs are attractive precisely because they might sidestep some of the tokamak's most expensive features — enormous superconducting magnet systems, in particular — but they generally rest on a thinner base of decades-long physics validation, and several aim explicitly at aneutronic or low-neutron fuel cycles (such as deuterium with helium-3, or proton-boron-11) that would sidestep the materials problem, at the cost of requiring even higher plasma temperatures that remain far from demonstrated.

None of these four bets is obviously correct, and it is worth resisting the temptation — common in press coverage — to declare a winner prematurely. Each faces a genuinely distinct, unresolved technical risk, and the field's history is littered with confident timelines that slipped by decades.

Because "commercial viability" is a moving target defined partly by economics rather than physics alone, there is no single experiment that will resolve this question the way, say, detecting a particular particle can resolve a question in high-energy physics. But there are concrete, checkable milestones that would mark genuine progress, each with its own rough — and appropriately hedged — timeline.

The nearest-term milestone is engineering breakeven in a magnetic device: a tokamak or stellarator that produces more fusion energy than the total electrical energy fed into its heating and magnet systems, not merely the energy delivered to the plasma. ITER is designed to demonstrate a fusion gain Q of 10, which if achieved would represent a genuine engineering breakeven with substantial margin, though ITER itself will not generate electricity — it has no turbines. That milestone is targeted for the 2030s, contingent on the current construction schedule holding, which recent history gives real reason to doubt; the project's timeline has slipped multiple times since its founding treaty in 2006.

A second, more commercially relevant milestone is a compact device — SPARC, or a competitor — demonstrating net energy gain from the plasma at high field in a machine small enough to plausibly template a fleet of power plants. Commonwealth Fusion Systems has publicly targeted operation of SPARC in the mid-to-late 2020s, with net energy gain (Q greater than 1 from the plasma) as an initial goal, to be followed by a demonstration power plant, ARC, in the early 2030s — an aggressive timeline that the company itself acknowledges carries execution risk, and that outside plasma physicists have both praised for ambition and questioned on confinement-scaling grounds.

A third milestone, perhaps the least glamorous but arguably the most decisive, is materials qualification: a dedicated high-flux 14 MeV neutron source that can irradiate candidate structural materials — reduced-activation ferritic-martensitic steels, vanadium alloys, silicon carbide composites — to fluences representative of decades of reactor operation, within a feasible test campaign of a few years. Facilities like IFMIF-DONES, under construction in Spain, are meant to provide exactly this capability; its timeline points toward the 2030s for meaningful data. Until such data exists, every claim about a fusion reactor's operating lifetime rests on extrapolation from lower-flux tests and computational models, not direct measurement — an honest gap that responsible fusion engineers acknowledge openly.

A fourth milestone is a demonstrated, self-sufficient tritium breeding cycle in an actual burning plasma environment — something no existing or currently under-construction machine, including ITER, is designed to fully demonstrate at reactor-relevant tritium breeding ratios, since ITER itself will test breeding blanket modules but is not expected to be tritium self-sufficient.

And the final, ultimate milestone is simply this: a fusion plant connected to a national electricity grid, delivering power at a levelized cost per kilowatt-hour that a utility commission judges competitive, sustained over a period — likely a decade or more — long enough to validate reliability, maintenance costs, and component lifetimes under real operating conditions rather than idealized test campaigns. Multiple government and industry roadmaps — the U.S. Department of Energy's "bold decadal vision," the UK's STEP program targeting the 2040s, China's ambitious state-led fusion timeline — now name the 2040s as a plausible window for a first grid-connected pilot plant. Every fusion scientist who has lived through the field's history treats such dates with appropriate caution: the joke that "fusion is thirty years away, and always will be" persists precisely because it has, for seventy years, been approximately true. What has genuinely changed in the past decade is that the joke is now measurably less accurate than before — real net energy gain has been demonstrated, private capital exceeding $8 billion has entered the field since 2021, and high-temperature superconductor technology has matured enough to shrink credible reactor designs substantially. Whether that acceleration is enough to finally break the pattern is, honestly, an open bet.

It is worth being precise about why this particular engineering problem carries such disproportionate cultural and scientific weight, beyond the simple appeal of "unlimited clean energy," a phrase that undersells both the difficulty and the nuance of what fusion would actually offer.

Fusion power, if achieved at scale, would not eliminate the need for a well-managed grid, storage, or transmission infrastructure — no single technology does that. But it would offer something distinctive: a dispatchable, high-density, carbon-free power source requiring no long-duration storage to smooth out weather, unlike solar and wind, and generating no long-lived high-level nuclear waste, unlike fission — although fusion's structural materials do become neutron-activated and require handling and eventual disposal, a lesser but real waste stream, typically decaying to low-hazard levels within decades to a century depending on material choice, rather than the tens of thousands of years associated with some fission waste products. Fuel supply is a near-non-issue: the deuterium in Earth's oceans alone represents an energy reserve that dwarfs any conceivable demand for millions of years, a scale that reframes energy scarcity as a solved problem in a way no fossil or even fission resource base can match, given uranium's more limited terrestrial abundance.

There is also a deeper, less transactional stake. Fusion sits at a genuine convergence point of physics frontiers rather than standing alone. Its magnet systems depend on advances in high-temperature and prospectively room-temperature superconductivity, where a material that could carry lossless current without cryogenic cooling would transform reactor design as thoroughly as REBCO tape already has. Its plasma diagnostics and control systems increasingly borrow from machine learning techniques developed for entirely different domains — DeepMind's collaboration with the Swiss Plasma Center in 2022 used reinforcement learning to control the shape of a tokamak plasma in real time, a small but telling sign of how adjacent computational frontiers are bleeding into fusion engineering. And some of the most exotic materials proposed for handling neutron damage and heat flux — including certain ordered alloys and structured composites — draw on the same broad materials-physics toolkit being developed for topological and quantum materials research, even though the connection there is looser and more exploratory than the superconductor link. Fusion, in other words, is not an isolated engineering slog; it is a proving ground where several of physics's live frontiers must simultaneously deliver.

Finally, there is the plain civilizational stake. Climate change is, at bottom, an energy-supply problem: humanity needs enormous quantities of reliable power without proportional carbon emissions, on a timescale of decades. Whether fusion arrives in time to matter for that transition — the 2040s and 2050s, on current best-guess roadmaps — or arrives too late to be more than a supplementary technology in a grid already decarbonized by other means, is itself an open and consequential question, one that depends as much on the pace of solar, storage, and fission deployment in the interim as on fusion's own progress.

Fusion research has always been an unusually international, unusually patient scientific culture — a field where careers span entire lifetimes without seeing the payoff their early work aimed at, and where collaboration across Cold War lines began remarkably early.

physicist[Lev Artsimovich|1909–1973|Soviet physicist who led the development of the tokamak at the Kurchatov Institute; the T-3 tokamak's 1968 results, showing plasma temperatures far higher than Western scientists initially believed possible, were confirmed by a visiting British team and triggered the tokamak's worldwide adoption as the leading confinement concept]

history 1920 — Arthur Eddington proposes stellar fusion as the sun's energy source 1951 — Lyman Spitzer proposes the stellarator; Argentina falsely claims fusion success in the Huemul Project 1968 — Soviet T-3 tokamak results confirmed by British team, triggering global tokamak adoption 1997 — JET achieves 16 MW of fusion power and 69 MJ over five seconds using deuterium-tritium fuel 2022 — NIF achieves scientific energy gain, laser-in to fusion-out, for the first time

Stand back from the individual machines and a striking pattern emerges: fusion is a field where the deepest physics — how nuclei overcome the Coulomb barrier, how a hot plasma can be confined by magnetic fields, how the Lawson criterion sets the terms of the trade — has been essentially settled since the mid-twentieth century, while nearly everything downstream of that physics remains genuinely, honestly open. That is an unusual and interesting shape for a scientific frontier to take. It is not that we lack a theory; it is that translating a well-understood theory into a durable, economical machine has proven to be its own species of hard problem, one where plasma turbulence, neutron-damaged steel, superconducting tape, and utility economics all have equal votes.

What makes the question gripping, rather than merely technical, is that it sits exactly at the boundary the essay opened with: a demonstrated scientific fact — fusion releases net energy — pressed up against an unresolved engineering and economic reality, with genuine intellectual disagreement among serious experts about which path, and which timeline, will actually get there. No one credible claims fusion power is impossible. Almost no one credible claims it is imminent in the way 1950s optimists once did. The honest position, held by most working plasma physicists, is that the physics case has been made, the materials and economic case has not, and the next twenty years of hard, unglamorous engineering — not a single eureka moment — will decide it.

Sub-questions still wide open:

• Can any structural material survive decades of 14 MeV neutron bombardment at commercially relevant fluence, and can we even test this before building a full reactor?
• Can a fusion power plant achieve a tritium breeding ratio reliably above 1.0 in practice, closing the fuel cycle without external supply?
• Do compact, high-field tokamak designs preserve favorable confinement scaling at the more extreme parameters they require, or do new instabilities emerge?
• Can inertial fusion lasers reach the repetition rates and wall-plug efficiencies a power plant demands, alongside affordable mass-produced targets?
• Will private-sector timelines, some promising net electricity by the early-to-mid 2030s, survive contact with the same engineering delays that have slowed every major public fusion project to date?
• Is there a fundamentally different confinement or fuel-cycle concept — aneutronic fusion, novel pinch geometries — waiting to leapfrog the whole tokamak-versus-stellarator-versus-laser framework entirely?

Ask the tutor: If a compact tokamak like SPARC does achieve net energy gain from its plasma later this decade, what specific engineering and materials hurdles would still stand between that result and an actual power plant delivering electricity to a grid — and how would you go about estimating, from public data, whether its projected economics are realistic?