Fusion Energy After Ignition: The Engineering Problems That Remain

In December 2022, a laser facility in California announced that a fusion experiment had produced more energy than the laser light delivered to its target. It was a genuine milestone, correctly celebrated, and almost universally misreported. Understanding precisely what was achieved — and what was not — is the clearest available introduction to why fusion power remains difficult.

What breakeven measured

Fusion releases energy when light nuclei combine. The reaction easiest to exploit on Earth uses deuterium and tritium, two hydrogen isotopes, which fuse into helium and a fast neutron and release about 17.6 megaelectronvolts per event. That is roughly four times the energy per nucleon of fission and millions of times that of a chemical bond. The obstacle is electrostatic: both nuclei are positively charged and repel each other, so they must be brought to conditions of extreme temperature and density before the strong nuclear force can take over.

The 2022 result concerned target gain: the fusion energy released divided by the laser energy that arrived at the fuel capsule. That ratio exceeded one. The number that matters for a power plant is different and much less favourable. Producing those laser pulses required roughly a hundred times more electrical energy than the light that reached the target, because the flashlamp-pumped laser architecture in question was designed for precision and diagnostic access, not efficiency. Add the plant's own consumption — cryogenics, vacuum systems, magnets, cooling, control — and engineering gain, the ratio of net electricity out to electricity in, remained far below one.

This is not a criticism of the experiment. The facility was built for weapons-physics research, not power generation. But it means the honest summary is: the physics of ignition has now been demonstrated; the engineering of a power plant has not begun in earnest.

Fusion's remaining difficulty is no longer principally about making the reaction happen. It is about making a machine that survives making it happen, thousands of times a day, for decades.

Two families of approach

Magnetic confinement holds a hot, diffuse plasma in place with magnetic fields. Because charged particles spiral along field lines, a closed toroidal field can keep plasma away from material walls. The tokamak achieves this with a strong toroidal field plus a current driven through the plasma itself; the stellarator achieves it with elaborately shaped external coils and no driven plasma current. Tokamaks are simpler to build and have delivered the best confinement results; stellarators are harder to design — modern ones exist only because optimisation on supercomputers made their coil geometry tractable — but avoid the current-driven instabilities that can abruptly terminate a tokamak discharge.

Typical operating conditions are counterintuitive: plasma temperatures above 100 million kelvin at densities roughly a million times thinner than air. There is very little material present, which is why the vessel is not instantly destroyed, but the energy flux to the region where field lines intersect the wall — the divertor — can approach ten megawatts per square metre. That is comparable to a rocket engine throat, sustained.

Inertial confinement takes the opposite route: compress a millimetre-scale fuel capsule so violently that it fuses before it can fly apart. Densities reach many times that of solid lead for a fraction of a nanosecond. The engineering challenges are correspondingly different — pulse-repetition rate, target manufacturing cost, and optics survival replace steady-state confinement and current drive.

A third category, magnetised target fusion, compresses a magnetised plasma mechanically on millisecond timescales, aiming for a middle ground in density and duration.

The high-field magnet shift

The most consequential recent change in magnetic fusion is not a plasma result. It is a magnet.

Fusion power density in a tokamak scales roughly with the fourth power of magnetic field strength. Doubling the field therefore permits, in principle, comparable performance in a machine of far smaller volume — and volume drives cost. Conventional low-temperature superconducting magnets, using niobium-tin or niobium-titanium, are limited to fields around 12 to 13 tesla at the coil in practical designs.

Rare-earth barium copper oxide tape, a high-temperature superconductor, changes those limits. It carries very high current density in strong fields and can operate at 20 kelvin rather than 4, which relaxes cryogenic engineering. In 2021, a large-bore magnet built from this material demonstrated a field above 20 tesla, and that demonstration is why several privately funded compact tokamak projects became credible enough to attract serious capital.

The remaining questions about these magnets are unglamorous and important: mechanical stress management, because magnetic pressure at 20 tesla exceeds 150 atmospheres; quench protection, since a local loss of superconductivity in a tape-wound coil dissipates stored energy in a small volume; and radiation tolerance of the tape and its insulation over years of neutron exposure.

The tritium problem

Deuterium is abundant — about one hydrogen atom in 6,400 in seawater. Tritium is not. It is radioactive with a half-life of 12.3 years, so it does not accumulate naturally. The world's civil inventory, produced largely as a byproduct in heavy-water fission reactors, is measured in tens of kilograms and is declining as those reactors retire.

A deuterium-tritium power plant must therefore breed its own tritium. The mechanism is a lithium blanket surrounding the plasma: fusion neutrons interact with lithium-6 to produce helium and tritium. Because each fusion event consumes one tritium and produces only one neutron, and because some neutrons are inevitably absorbed by structural material or escape, the blanket must include a neutron multiplier such as beryllium or lead, and the tritium breeding ratio must exceed one with margin.

No facility has yet demonstrated a closed fuel cycle at power-plant scale. The engineering subproblems are substantial. Tritium is the smallest and most mobile isotope of the smallest element; it permeates hot metals, so containment and accounting are genuinely hard. It must be extracted continuously from the breeder material, separated from other hydrogen isotopes, purified, and reinjected. Inventory must be minimised for both safety and economics, since tritium is extraordinarily expensive.

Alternative fuel cycles avoid tritium but pay elsewhere. Deuterium-deuterium fusion needs substantially higher temperatures and still produces tritium as an intermediate. Proton-boron-11 fusion produces charged particles rather than neutrons, which is attractive for materials lifetime and direct energy conversion, but requires conditions well beyond anything demonstrated, with severe bremsstrahlung radiation losses.

Materials under neutron flux

The 14 MeV neutron from deuterium-tritium fusion is the plant's central materials problem. It carries far more energy than a typical fission neutron and it is electrically neutral, so magnetic fields do not deflect it. It passes into the structure and causes two distinct kinds of damage.

The first is displacement damage: atoms knocked out of lattice positions, creating cascades of vacancies and interstitials. Over years, this hardens and embrittles steel, reducing fracture toughness, and can cause void swelling and dimensional change.

The second is transmutation. High-energy neutrons trigger reactions that produce helium and hydrogen inside the metal. Helium is insoluble; it collects at grain boundaries and forms bubbles, which drives embrittlement more aggressively than displacement damage alone. Transmutation also creates activated isotopes, which is why reduced-activation ferritic-martensitic steels — engineered to avoid elements that produce long-lived activation products — are the leading structural candidates. Silicon carbide composites, tungsten armour, and vanadium alloys occupy various niches.

A frustrating aspect of this work is testing. There is currently no operating high-flux 14 MeV neutron source dedicated to qualifying fusion materials at power-plant lifetime doses. Accelerator-based irradiation facilities intended to fill this gap are under construction and remain years from producing full datasets. Meanwhile, designers must extrapolate from fission-spectrum irradiation and ion-beam surrogates, both imperfect analogues.

From heat to electricity

An often-skipped step: most fusion concepts generate electricity the same way a coal plant does. Neutrons deposit energy as heat in the blanket, a coolant carries that heat away, and it drives a turbine through a thermodynamic cycle. Conversion efficiency is governed by the coolant outlet temperature, which pushes designers toward helium or molten salt at high temperature rather than water — and every increase in temperature intensifies the materials problem.

This is where the plant-level arithmetic bites. Suppose a tokamak achieves a plasma gain of 10, meaning fusion power ten times the power injected to heat the plasma. Roughly 80 percent of that fusion energy leaves as neutron kinetic energy and is captured as heat; thermal conversion returns perhaps 35 to 40 percent as electricity. From that gross electrical output you must subtract heating-system inefficiency, magnet cryogenics, pumping power, tritium plant load, and balance-of-plant consumption. Positive net output requires either substantially higher gain or unusually efficient recirculating systems. This is why credible plant designs target gains well above the threshold needed for scientific breakeven.

Then there is availability. A power plant earns revenue only while running. Fusion machines have historically operated in short pulses with long maintenance intervals, and remote maintenance inside an activated vessel — where human entry is impossible — is a robotics problem of unusual severity. Replacing a divertor cassette in an activated tokamak may take weeks. A plant that runs 50 percent of the year competes very differently from one that runs 90 percent.

Where fusion's advantages are real

None of this makes fusion unattractive. Its safety and waste characteristics are genuinely different from fission, and the differences are structural rather than procedural.

There is no chain reaction to run away: the plasma is a fragile state that extinguishes when confinement is lost. Fuel inventory in the reaction chamber at any instant is a fraction of a gram, so there is no large stored chemical or nuclear energy release available. Activated structural material is a real waste stream, but with appropriate material choices its radiotoxicity decays over roughly a century rather than hundreds of thousands of years, and it does not include long-lived actinides. Fuel supply is effectively unlimited and geographically distributed. Output is dispatchable thermal generation, which complements variable renewables in a way that additional variable generation does not.

A reasonable expectation

Public and private programmes now pursue several parallel routes: a large international tokamak experiment intended to demonstrate sustained burning plasma at scale, compact high-field tokamaks aiming for faster iteration, optimised stellarators pursuing steady-state operation, and inertial and magnetised-target concepts with different repetition-rate challenges.

A defensible reading of the evidence is that demonstrating net electricity to a grid from a fusion device is plausible within the 2030s for at least one programme, and that a fleet of plants competing on cost is a 2040s-and-beyond proposition — gated less by physics than by supply chains for superconducting tape, tritium handling experience, qualified structural materials, and the operational learning that only comes from running machines for years.

The useful mental model is not "fusion is twenty years away, always." It is that fusion has moved from a physics question to an industrial one. Industrial problems are, historically, the kind humanity solves — slowly, expensively, and then all at once.