Nuclear fusion projects
Fusion is not on this site's fleet statistics because no fusion device has ever produced net electricity, and none is connected to a grid. What follows is what each machine has actually demonstrated, kept separate from what it is aiming at — a distinction that most fusion coverage collapses.
The two numbers that get confused
Scientific gain compares the fusion energy released against the energy delivered to the fuel. NIF passed this in December 2022: just over 2 MJ of laser energy into the target, 3.15 MJ of fusion energy out. It was a real milestone and it is repeatedly reported as "fusion produces more energy than it consumes".
Engineering gain compares electricity out against electricity in for the whole facility. NIF's lasers draw roughly 300 MJ from the grid to deliver those 2 MJ. On that measure no device has come close, and that is the measure a power station is judged by.
Fusion's fuel is deuterium and tritium. Deuterium is abundant — 30 grams in every cubic metre of seawater. Tritium is the constraint: it is radioactive with a 12-year half-life, occurs naturally only in traces, and the entire world civilian stock is roughly 25 kilograms.
Almost all of it is Canadian, and it is a byproduct of fission. A CANDU heavy water reactor produces about 130 grams of tritium a year in its moderator, and Ontario Power Generation extracts it at the Darlington Tritium Removal Facility. ITER alone expects to consume about 12 kg over its operating life; a single commercial plant's startup inventory could be 10 kg.
So the machine that fusion depends on for its startup fuel is a fission reactor — and specifically the one type that needs no enrichment at all. Beyond startup, every design intends to breed its own tritium from lithium in a blanket surrounding the plasma. No blanket has ever done this at scale, and tritium breeding is arguably a harder unsolved problem than the plasma physics.
Every project
The "demonstrated" column is what a machine has actually done, with its date. Targets are shown separately, and never as achievements.
| Project | Approach | Country | Status | Target |
|---|---|---|---|---|
| EAST Institute of Plasma Physics, Chinese Academy of Sciences | tokamak | China | — | |
| JT-60SA QST and EUROfusion | tokamak | Japan | — | |
| KSTAR Korea Institute of Fusion Energy | tokamak | South Korea | — | |
| Laser Mégajoule CEA | inertial | France | — | |
| LM26 General Fusion | magnetised target | Canada | — | |
| National Ignition Facility Lawrence Livermore National Laboratory | inertial | United States | — | |
| Polaris and Orion Helion Energy | FRC | United States | 2028, 50 MW | |
| ST40 Tokamak Energy | tokamak | United Kingdom | — | |
| Wendelstein 7-X Max Planck Institute for Plasma Physics | stellarator | Germany | — | |
| ITER ITER Organization (35 nations) | tokamak | France | Under construction | 2034, 500 MW |
| SPARC Commonwealth Fusion Systems | tokamak | United States | Under construction | 2027, 100 MW |
| ARC Commonwealth Fusion Systems | tokamak | United States | Planned | early 2030s, 400 MW |
| Infinity Two Type One Energy | stellarator | United States | Planned | 2028, 400 MW |
| K-DEMO Korea Institute of Fusion Energy | tokamak | South Korea | Planned | 2037 |
| STEP UK Industrial Fusion Solutions / UKAEA | tokamak | United Kingdom | Planned | 2040 |
| JET (Joint European Torus) UKAEA / EUROfusion | tokamak | United Kingdom | Closed | — |
The approaches
- Tokamak
- A doughnut of plasma confined by magnetic fields, with a current driven through the plasma itself. The most developed approach by far, and the one ITER and most private ventures use. Its weakness is that the plasma current can disrupt violently.
- Stellarator
- The same doughnut, but the twist comes from the shape of the coils rather than a plasma current. Far harder to build — the magnets are extraordinary shapes — but no disruptions, and it can in principle run continuously.
- Inertial confinement
- Crush a fuel pellet with lasers so fast that it fuses before it can fly apart. NIF is the flagship. A power plant would need to do this several times a second, indefinitely; NIF fires roughly once a day.
- Field-reversed configuration and magnetised target
- Pulsed approaches that compress a plasma mechanically or magnetically, aiming to skip the steam cycle and take electricity directly. Helion and General Fusion are the main proponents. Less proven physics, but potentially far smaller machines.