A fusion generator is a machine designed to produce electricity by fusing light atomic nuclei together, releasing energy from the same process that powers the sun. No fusion generator feeds electricity into a power grid today, but the physics of energy release has been demonstrated in the laboratory, and dozens of public and private programs are now engineering toward commercially viable designs. The gap between a physics demonstration and a working power plant is still wide, filled with unsolved problems in materials science, fuel supply, plasma stability, and cost, but each of those problems has seen measurable progress in recent years.
How Fusion Releases Energy
When two light nuclei merge into a heavier one, the resulting nucleus weighs slightly less than the sum of its parts. That missing mass becomes energy. The reaction most fusion projects pursue combines deuterium and tritium, two heavy forms of hydrogen, to produce a helium nucleus and a fast neutron. The neutron carries most of the energy, roughly 14 million electron-volts per reaction, which is millions of times more energy per unit mass than any chemical fuel can deliver.
Getting nuclei to fuse is the hard part. Both carry a positive charge, so they repel each other. To overcome that repulsion, the fuel must be heated to temperatures above 100 million degrees, far hotter than the core of the sun. At those temperatures the fuel exists as a plasma, a gas-like state where electrons have been stripped from their atoms. Keeping that plasma hot and dense long enough for a useful number of fusion reactions to occur is the central challenge. Physicists express this challenge through what is known as the Lawson criterion: the product of plasma density, confinement time, and temperature must exceed a threshold for the energy coming out to surpass the energy going in. For breakeven, that triple product needs to exceed roughly 10²¹ keV·s·m⁻³ at temperatures of tens of keV.1Frontiers in Nuclear Engineering. Evaluation of the Lawson criterion for aneutronic proton-boron-11 fusion: effects of ion temperature and bremsstrahlung losses
Two Roads to Confinement
Every fusion generator concept has to solve the confinement problem: how do you hold a plasma at 100 million degrees without it touching anything? Two broad strategies dominate the field, and they look nothing alike.
Magnetic Confinement
The first approach uses powerful magnets to trap the plasma in a shaped magnetic bottle. The most familiar design is the tokamak, a doughnut-shaped chamber where magnetic field lines spiral around the ring, keeping charged particles from drifting into the walls. ITER, the international experiment under construction in southern France, is a tokamak. A less well-known cousin is the stellarator, which achieves the same confinement through a more complex, twisted coil geometry rather than relying on an electric current flowing through the plasma itself. Recent work has explored stellarator designs using arrays of simpler planar coils to produce the required magnetic fields. One optimized concept targets a two-field-period quasi-axisymmetric configuration with an on-axis field of 6 T and a major radius of 2.7 m.2IOP Publishing (Nuclear Fusion). Stellarator fusion systems enabled by arrays of planar coils The advantage of stellarators is that they can, in principle, run continuously without the pulsed operation that limits some tokamak designs.
Inertial Confinement
The second approach skips magnets entirely. Instead, it crushes a tiny pellet of fuel so fast that the fuel’s own inertia holds it together long enough for fusion to occur. The most dramatic example is the National Ignition Facility (NIF) in the United States, which fires 192 laser beams at a target to compress deuterium-tritium fuel. In August 2021, the NIF produced 1.37 MJ of fusion energy in a single shot using an indirect-drive approach, where laser light heats a small metal capsule that then radiates X-rays onto the fuel pellet.3PubMed. Design of an inertial fusion experiment exceeding the Lawson criterion for ignition A subsequent shot in December 2022 went further, producing more fusion energy than the laser delivered to the target. One of the persistent headaches for inertial confinement is the Rayleigh-Taylor instability, where tiny imperfections in the pellet’s surface get amplified during compression, causing the fuel to mix unevenly. Careful tuning of laser pulse shapes and target structures can substantially reduce this problem; one study found that optimized profiles increased areal density by over 68% compared to an unoptimized baseline.4Nuclear Fusion. Achieving stable implosion through hydro-equivalent analysis for direct-driven inertial confinement fusion
Heating the Plasma
Even in a magnetic confinement device, the plasma does not heat itself to fusion temperatures spontaneously. External heating systems do the heavy lifting. One widely used method is neutral beam injection: fast atoms of deuterium or hydrogen are accelerated and fired into the plasma, where they collide with plasma particles and transfer their kinetic energy as heat. A tokamak-based design for a volumetric neutron source, for example, uses roughly 42 MW of deuterium neutral beam injection for heating, current drive, and to provide fast ions for fusion reactions.5Fusion Engineering and Design. Neutral Beam Injection for a tokamak-based Volumetric Neutron Source Other heating methods include radiofrequency waves tuned to resonances within the plasma, and direct ohmic heating from the electric current that circulates through a tokamak’s plasma ring. In a working fusion generator, the goal is to reach a state where the helium nuclei produced by fusion reactions deposit enough energy back into the plasma to keep it hot, reducing the need for external heating.
The Materials Problem
Building a container for a 100-million-degree plasma sounds impossible, and in a sense it is. The magnetic field does the confining, but the chamber walls still get bombarded by energetic neutrons, photons, and escaping plasma particles. The components facing the plasma must withstand extraordinary punishment. In ITER and similar designs, the divertor, the component at the bottom of the chamber that exhausts waste heat and helium ash, faces steady-state heat loads around 10 MW per square meter. During sudden plasma disruptions or instability events, those loads spike to between 0.6 and 3.5 GW per square meter for brief periods.6Physics of Atomic Nuclei. Results of high heat flux tests of tungsten divertor targets under plasma heat loads expected in ITER and tokamaks (review)
Tungsten is the leading candidate for plasma-facing surfaces because it has the highest melting point of any metal and good thermal conductivity. But testing reveals that even tungsten suffers. Under transient loads, the surface recrystallizes, cracks form, and molten layers move and erode, producing submicron-sized dust that can contaminate the plasma.7Nuclear Materials and Energy. Behavior of divertor and first wall armour materials at plasma heat fluxes relevant to ITER ELMs and disruptions Copper alloys back up the tungsten as coolant tube materials, valued for their thermal conductivity.8Nuclear Fusion. Baseline high heat flux and plasma facing materials for fusion
Beyond surface damage, the 14 MeV neutrons produced by deuterium-tritium fusion penetrate deep into structural materials and cause damage at the atomic level. One critical effect is helium production through nuclear transmutation reactions. Helium atoms are virtually insoluble in metals, so they clump into tiny bubbles along grain boundaries, promoting swelling and making the material brittle. Among the candidate structural alloys evaluated for compact fusion reactor designs, vanadium-chromium-titanium alloys and certain reduced-activation steels show the lowest helium and hydrogen generation rates, while nickel-based superalloys and silicon carbide composites generate considerably more.9Frontiers in Nuclear Engineering. Assessment of structural materials in compact fusion reactor design Helium effects can become the determining factor for how long a component lasts, particularly at high temperatures.10Nuclear Fusion. The influence of helium on the bulk properties of fusion reactor structural materials This means that the inner wall components of a fusion generator would need periodic replacement, a cost and downtime factor that looms large in economic projections.
The Tritium Fuel Challenge
Deuterium, one half of the fuel mix, is abundant in ordinary seawater and costs very little to extract. Tritium is a different story. It is radioactive with a half-life of about 12.3 years, meaning natural supplies are negligible. Today’s tritium comes mostly from Canadian CANDU heavy-water fission reactors as a byproduct, but the quantities are small. A fusion power plant producing 1,000 MW of fusion power would burn through roughly 56 kg of tritium per year.11Nuclear Fusion. Physics and technology considerations for the deuterium–tritium fuel cycle and conditions for tritium fuel self sufficiency By comparison, CANDU reactors produce only about 130 grams of tritium per gigawatt-electric per full-power year. The global inventory has been declining since about 2025 as demand and natural decay outpace supply.12Fusion Engineering and Design. Tritium supply and use: a key issue for the development of nuclear fusion energy
The plan for a working fusion generator is to breed its own tritium. The 14 MeV neutrons from fusion reactions would be captured in a “breeding blanket” surrounding the plasma chamber, where they strike lithium to produce fresh tritium. Achieving a tritium breeding ratio greater than one, meaning the reactor makes more tritium than it consumes, is essential for a self-sustaining fuel cycle. One blanket design uses a eutectic mixture of lead and lithium as the breeding material, with water as the coolant. Getting the breeding ratio close to one in practice requires careful engineering; studies have shown that adding beryllium as a neutron multiplier can help reach the needed production rate.13Fusion Engineering and Design. A water cooled, lithium lead breeding blanket for a DEMO fusion reactor Work on stellarator-specific blanket designs using dual-coolant lithium-lead concepts is also under way, with simplified models used to estimate variations in tritium breeding ratio and radiation damage for different first-wall configurations.14Energies. Neutronic Assessments towards a Novel First Wall Design for a Stellarator Fusion Reactor with Dual Coolant Lithium Lead Breeding Blanket No breeding blanket has yet been tested at the conditions of a full-power fusion reactor, so this remains one of the technology’s biggest open questions.
Keeping the Plasma Stable
A fusion plasma is not a well-behaved substance. It is prone to sudden collapses called disruptions, where the confined plasma dumps its energy into the chamber walls in milliseconds. A disruption in a large device like ITER could deliver enormous forces to the vessel structure and erode the divertor surface. Predicting and preventing disruptions is therefore a safety-critical engineering task, not just a physics curiosity.
Classical approaches to disruption prediction relied on monitoring a handful of plasma parameters and sounding an alarm when thresholds were crossed. More recently, deep learning models trained on data from major tokamaks, including DIII-D in the United States and the Joint European Torus (JET) in the UK, have delivered more reliable predictions with longer warning times. A particularly important feature of these models is their ability to generalize: a model trained on data from one machine can predict disruptions on a different machine it has never seen before.15Nature. Predicting disruptive instabilities in controlled fusion plasmas through deep learning That transferability matters because future reactors will be too expensive to use as training grounds for disruption avoidance. The shift from passive prediction to active, real-time control of plasma stability is one of the areas where machine learning is making the strongest near-term contribution to fusion engineering.
Safety and Radioactive Waste
Fusion is frequently described as “clean” energy, and while that reputation has some basis, it needs qualification. A deuterium-tritium fusion generator would not produce the long-lived transuranic waste (plutonium, americium) that makes fission waste dangerous for thousands of years. The radioactive waste from fusion comes mainly from neutron activation of the structural materials surrounding the plasma, and from tritium itself. When reduced-activation materials are used, the resulting waste shows relatively rapid radioactive decay and low decay heat, which simplifies handling.16Nuclear Fusion. Overview on the management of radioactive waste from fusion facilities: ITER, demonstration machines and power plants
Inventory calculations for the European DEMO reactor concept illustrate the timeline. The divertor, which absorbs the most neutron damage, would be classifiable as low-level waste within about a century of shutdown. The vacuum vessel, being a much larger and more heavily irradiated structure, would take longer to meet low-level waste criteria but could be potentially recyclable within a similar time frame.17Nuclear Fusion. Activation, decay heat, and waste classification studies of the European DEMO concept That is a vast improvement over fission waste, which requires secure storage for tens of thousands of years in the case of spent fuel.
Tritium itself is a safety concern. It is radioactive and can be absorbed biologically, so fusion facilities must minimize inventories and prevent releases. The quantities involved are significant, kilograms of tritium circulating through the fuel cycle, and any leak would need to be contained. Safety approaches center on minimizing the amount of tritium in the system at any given time, reducing tritium permeation through materials, and decontaminating waste.18Nuclear Fusion. Physics and technology considerations for the deuterium–tritium fuel cycle and conditions for tritium fuel self sufficiency A fusion reactor also cannot “melt down” in the fission sense: if confinement is lost, the plasma simply cools and the reactions stop. There is no chain reaction to run away.
Converting Fusion Energy to Electricity
Even once a fusion generator produces net energy, that energy still has to become electricity. In a deuterium-tritium reactor, about 80% of the energy leaves as fast neutrons. These neutrons are absorbed in the breeding blanket and converted to heat, which can then drive a conventional steam turbine. The thermal cycle efficiency of such a system would resemble that of a modern thermal power plant, somewhere in the range of 30 to 40 percent, depending on coolant temperatures.
Some advanced concepts aim to bypass the steam cycle altogether. Direct energy conversion schemes attempt to capture the kinetic energy of charged particles, the helium nuclei and unreacted ions escaping the plasma, and convert it directly to electricity. This is especially appealing for aneutronic fuels (fuels that produce few neutrons), where most of the energy comes out in charged particles rather than neutrons. However, achieving high-efficiency direct conversion appears feasible only in open magnetic field line configurations, which struggle with the plasma confinement needed for aneutronic fuels in the first place.19Plasma Physics and Controlled Fusion. Generic issues for direct conversion of fusion energy from alternative fuels Newer adiabatic direct conversion concepts using crossed electric and magnetic fields may sidestep some of the classical drawbacks of thermionic and magnetohydrodynamic devices, but they remain at an early research stage.20PRX Energy. Efficiency and Physical Limitations of Adiabatic Direct Energy Conversion in Axisymmetric Fields
Alternative Fuels Beyond Deuterium-Tritium
Deuterium-tritium fuel dominates current research because it has the lowest ignition temperature of any fusion reaction, but it is not the only option. Deuterium-helium-3 and proton-boron-11 are sometimes called “advanced” or aneutronic fuels because they produce far fewer neutrons, which would reduce activation damage and radioactive waste. The proton-boron reaction in particular is sometimes described as the “holy grail” of fusion fuels because its primary products are just three helium nuclei with almost no neutrons.21Physical Review Research. Proton-boron fusion scheme taking into account the effects of target degeneracy
The catch is that these reactions require vastly higher temperatures and confinement conditions. Boron’s higher nuclear charge means the plasma radiates away more energy as bremsstrahlung, X-ray light emitted when electrons are deflected by ions. Reaching net energy gain with proton-boron fuel demands extraordinarily dense and hot conditions that no current device comes close to achieving. These fuels also require much higher plasma beta (the ratio of plasma pressure to magnetic pressure) and much longer confinement times than deuterium-tritium, in a system of smaller size.22Plasma Physics and Controlled Fusion. Generic issues for direct conversion of fusion energy from alternative fuels For the foreseeable future, any fusion generator that reaches the grid will almost certainly burn deuterium-tritium.
What Would Fusion Electricity Cost
The economics of a technology that does not yet exist are inherently speculative, but modeling gives some useful guardrails. For magnetically confined fusion to compete with other low-carbon sources beyond the 2040s, its levelized cost of electricity would likely need to land at or below roughly $80 to $100 per megawatt-hour at 2020 prices. Early fusion plants, whether large or small, will probably come in above $150 per MWh because of low power availability from pulsed operation, the frequent need to replace irradiated vessel components, and relatively low thermal cycle efficiency.23Energy Policy. Can fusion energy be cost-competitive and commercially viable? An analysis of magnetically confined reactors Those numbers improve with technology maturation: better cycle efficiency, longer-lived components, and manufacturing standardization all push costs down over time.
Inertial fusion economics work differently. Because each shot produces a burst of energy rather than continuous output, cost-effectiveness depends heavily on how much energy each shot yields and how often shots can be repeated. Modeling suggests that a combination of very high gain (above 500) and high energy yield per shot (above 5 GJ) could unlock designs with costs as low as $25 per MWh under optimistic but not unrealistic assumptions.24Philosophical Transactions of the Royal Society A. A simplified economic model for inertial fusion Reaching those gain and yield numbers is an enormous physics and engineering challenge that remains unsolved.
The Public-Private Race and Competing Timelines
For decades, fusion was almost exclusively a government-funded endeavor. ITER, jointly funded by seven parties representing 35 nations, has been under construction since 2010 and aims for first plasma in the early 2030s. The ITER-to-DEMO pathway envisions commercial fusion around the middle of the century. That pace frustrates many observers, but it reflects the genuine difficulty of the engineering.
Over the past decade, private fusion companies have raised billions in venture capital, each promoting a proprietary approach and aggressive timelines. The dynamics of private investment shape these promises: venture capital typically seeks returns within three to five years and rarely commits to opportunities with cash flows beyond ten years. That pressure pushes start-ups to present optimistic schedules. Surveys of private fusion companies show that the vast majority claim fusion electricity will reach the grid by the 2030s, while the publicly funded mainstream approach targets mid-century.25Environmental Research: Energy. All ways to fusion: a controversy mapping of the public-private shift in nuclear fusion research Whether private speed or public caution proves more realistic remains one of the defining tensions in the field. The honest answer is that no one has yet built a fusion device that produces net electricity, let alone one that runs continuously and affordably, and the gap between a physics proof-of-concept and a grid-connected generator is where most of the hard work still lives.
Why Fusion Generators Still Do Not Exist on the Grid
Each section above describes one piece of the puzzle, but the difficulty of fusion becomes clearer when you see them all stacked together. You need a plasma hot and dense enough to sustain fusion. You need magnets or lasers powerful enough to confine or compress it. You need wall materials that survive years of neutron bombardment without cracking, swelling, or generating excessive waste. You need a breeding blanket that produces more tritium than the reactor consumes, tested at conditions no facility can yet provide. You need a power conversion system efficient enough for the economics to work. And you need all of these systems to function simultaneously, reliably, for months or years at a stretch. Solving any one of these problems in isolation is impressive; solving them all together in a single integrated machine is the engineering grand challenge of fusion energy. The progress is real, but so is the distance remaining.