№ 124 · physics

Fusion: where it actually stands, not the slogan

A reader asked where fusion actually stands, without the slogan. The short answer: a laser lab in California has made more fusion energy than its laser put onto the fuel, at least eleven times by June 2026, and its best shot gave back about four times what went in. No fusion machine has reported making more energy than it drew from the grid, let alone sending out more electricity than it uses. The largest magnetic machine, ITER, plans to start burning the real fuel in 2039.

grid 100net -98.8-1000+200+400heat 0.60x0.012.01110040%Q 4.13.011100
4.13
0.6%
1 · fusion out / heating in2.48 / 0.60 = 4.13clears
2 · electricity back / grid draw1.23 / 100 = 0.012short
3 · net to the grid-98.8 per 100 drawnshort

Wall-plug gain 0.025. The fuel gave back more than it was heated with, and the plant still loses.

Blue: energy flowing round the loop; red: energy lost as heat at the heater and the turbine. Pipe fill width is on a log scale, like the widget's barsGauges, red tick at 1: fusion out over heating in (on the chamber), electricity back over grid draw (on the return pipe), and net to the grid, red tick at 0 (on the cabinet)Turbine fixed at 40%. Magnets, cooling and pumps are left out, which flatters every case
Pick a preset or turn the chamber's gain knob and the heater valve, and follow 100 units of grid electricity round the loop to the three gauges. Same arithmetic, presets and numbers as the widget below; pipe widths and motion are illustrative.

What has to happen

Fusion power plans almost all use two heavy forms of hydrogen, deuterium and tritium. When a pair of their nuclei fuse they make a helium nucleus and a free neutron and release 17.6 MeV: about 3.5 MeV goes to the helium and 14.1 MeV to the neutron. The hard part is getting them to touch. Both nuclei are positively charged and push each other away, so the fuel has to be a plasma at upwards of a hundred million degrees, and it has to stay hot and dense long enough for enough pairs to fuse to repay the heating. In 1955 John Lawson worked out that condition. Its usual modern form multiplies density, temperature and the time the heat stays in, the triple product, and asks whether it clears a threshold.

There are two main ways to try. Magnetic confinement holds a thin plasma for seconds inside magnetic fields, most often in a doughnut-shaped machine called a tokamak. Inertial confinement crushes a peppercorn-sized fuel capsule so fast that it fuses before it can fly apart; at the National Ignition Facility (NIF), up to 192 laser beams do the crushing.

Three different things called breakeven

Most of the slogan lives in one word: gain. Samuel Wurzel and Scott Hsu, in a review that compiles published results from fusion experiments across five decades, separate three kinds. Scientific gain is the fusion energy out divided by the heating energy delivered to the plasma or the target. Wall-plug gain divides the fusion energy by the electricity drawn from the grid to heat, hold and control the fuel. Engineering gain compares the electricity a plant sends to the grid with the electricity it feeds back into itself to keep going. They call a wall-plug gain of 1, not a scientific gain of 1, fusion's "Kitty Hawk moment".

The gap between the gains is efficiency. In their table of typical values, the heaters of a magnetic machine turn about 70% of their electricity into heating power, a laser built for a power plant about 10%, and a steam plant turns about 40% of the heat back into electricity. Their rule of thumb for practical power is that scientific gain times heater efficiency should reach about 10. Taking heaters at a cautious 50% or better for magnetic machines, they put the need at a scientific gain of about 20; for lasers, at about 100.

Interactive Drag the sliders (or tab to one and use the arrow keys), or pick a preset. Follow 100 units of electricity from the grid through the heaters, the fuel and the steam plant, and see what comes back.

4.13
0.6%
40%

Wall-plug gain 0.025: the fusion energy is 2.48 for every 100 units drawn from the grid. Electricity back 1.23 of 100. Net loss of 98.8. The fuel gave back more than it was heated with, and the plant still loses.

Our own drawing, not a figure from any source. The arithmetic is the one Wurzel and Hsu (2022) use: heating = grid electricity × heater efficiency; fusion = heating × scientific gain; electricity back = (fusion + heating) × heat-to-electricity efficiency, which they take as about 40% for a steam plant, with the extra heat a lithium blanket adds folded in. The black line on each bar marks the 100 units that came in. Presets: NIF's 7 April 2025 gain of 4.13 with a heater efficiency of 0.6%, roughly what a 2.05 MJ shot drawing about 300 to 400 MJ from the grid implies (we use the first ignition shot's grid draw; LLNL has not published one for the record shot). JET's 2023 gain of 0.37 and ITER's goal of 10 are paired with the review's typical 70% for magnetic-machine heaters, not with either machine's real heaters, and ITER will not make electricity at all. A real plant also spends power on magnets, cooling and pumps, which this leaves out, so it flatters every case.

Where the records stand

On 5 December 2022, NIF's lasers delivered 2.05 MJ to a target and got 3.15 MJ of fusion energy back, the first scientific gain above 1 in any controlled fusion experiment. The record since came on 7 April 2025: 8.6 MJ from 2.08 MJ, a gain of 4.13. The laser itself was not built to be efficient. For that first ignition shot it drew about 300 to 400 MJ from the grid, Jean-Michel Di Nicola, NIF's laser systems chief engineer, said in 2023, so if a later shot drew about the same, even 8.6 MJ is only a few percent of the electricity that went in. LLNL presents ignition as serving its nuclear weapons stockpile work as well as the prospect of fusion energy.

For tokamaks, the 2025 update of the review still lists JET's 1997 shot as the highest scientific gain, about 0.6 by its reckoning (0.67 is the figure usually quoted). JET's last deuterium-tritium campaign, in late 2023, set an energy record of 69 MJ over about five seconds from 0.2 milligrams of fuel, at a scientific gain of 0.37. JET ended its experiments that December. The two approaches trade differently: tokamaks hold heat for seconds, the laser burns harder but only for an instant.

What comes next, and when

ITER, the international tokamak being built in France, rewrote its schedule in 2024: research operation with hydrogen and deuterium from 2034, full magnetic energy in 2036, and deuterium-tritium fuel from 2039, four years later than the plan it replaced. Its Council endorsed that approach in November 2024. ITER's stated goals are 500 MW of fusion power from 50 MW of heating, a scientific gain of 10, and pulses of 400 seconds. Producing electricity is not among them.

Private companies promise sooner. Commonwealth Fusion Systems, building a compact tokamak called SPARC with high-temperature superconducting magnets, said in May 2026 that it expects to begin operating the machine in 2027, aiming for a scientific gain above 1, and in August 2026 that SPARC was almost 80% complete. That is the company's own timeline.

The fuel the slogan skips

Tritium is radioactive, with a half-life of 12.3 years, and scarce. Rachel Lawless, who heads tritium fuel cycle research at the UK Atomic Energy Authority, put the world's supply at about 30 kg in an IAEA webinar in December 2025, and said a 2.5 GW plant would burn 0.4 kg a day. So a power plant would have to make its own. The plan is a blanket of lithium around the plasma: the neutrons, which carry 80% of the energy, strike the lithium and make new tritium, deposit their heat, and are kept off the outer machine. The blanket has to make more than one tritium for each one burned, and that still has to be shown in a fusion machine: testing it is one of the jobs of ITER's test blanket programme, which in September 2026 picked four designs to try during its deuterium-tritium phase.

In one breath

Fusion has passed one real milestone: a laser shot that made more fusion energy than the laser delivered, now four times more. It has not passed the ones that make power: more energy than drawn from the wall, more electricity out than in, and fuel bred faster than it burns. The big public machine plans real fuel in 2039; a private one says 2027.

Where this comes from

  1. Progress toward fusion energy breakeven and gain as measured against the Lawson criterion linked only, not reproduced
    Samuel E. Wurzel and Scott C. Hsu · arXiv:2105.10954 · 2021
    arxiv.org/abs/2105.10954
  2. Continuing progress toward fusion energy breakeven and gain as measured against the Lawson criteria linked only, not reproduced
    Samuel E. Wurzel, Scott C. Hsu · arXiv:2505.03834 · 2025
    arxiv.org/abs/2505.03834
  3. Achieving Fusion Ignition (record of ignition shots, 2022 to June 2026) linked only, not reproduced
    Lawrence Livermore National Laboratory, NIF & Photon Science · 2026
    lmf.llnl.gov/science/achieving-fusion-ignition
  4. IAEA Webinar Explores NIF's Ignition and Energy Gain Breakthroughs linked only, not reproduced
    Charlie Osolin (LLNL) · 2023
    lmf.llnl.gov/news/iaea-webinar-explores-nifs-ignition-energy-gain-breakthroughs
  5. Fusion research facility JET's final tritium experiments yield new energy record linked only, not reproduced
    UK Atomic Energy Authority and EUROfusion · 2024
    www.ukaea.org/news/fusion-research-facility-jets-final-tritium-experiments-yield-new-energy-record/
  6. New baseline to prioritize robust start to exploitation linked only, not reproduced
    ITER Organization · 2024
    www.iter.org/node/20687/new-baseline-prioritize-robust-start-exploitation
  7. Support for baseline approach (35th ITER Council) linked only, not reproduced
    ITER Organization · 2024
    www.iter.org/node/20687/support-baseline-approach
  8. ITER in a few lines linked only, not reproduced
    ITER Organization · 2025
    www.iter.org/few-lines
  9. CFS begins preparing a home for our 100 million degree fusion plasma linked only, not reproduced
    Alex Graziano (Commonwealth Fusion Systems) · 2026
    blog.cfs.energy/cfs-begins-preparing-a-home-for-our-100-million-degree-fusion-plasma
  10. SPARC progress, DOE Milestone approvals reflect CFS power plant maturity linked only, not reproduced
    Jennifer Ganten (Commonwealth Fusion Systems) · 2026
    blog.cfs.energy/sparc-progress-doe-milestone-approvals-reflect-cfs-power-plant-maturity/
  11. Global Approaches to Tritium Breeding for Fusion: Materials, Challenges, and Pathways linked only, not reproduced
    Rachel Lawless, Giacomo Aiello, Luciano Giancarli & Richard Kamendje (IAEA webinar) · 2025
    www.iaea.org/sites/default/files/global-approaches-to-tritium-breeding-for-fusion-materials-challenges-and-pathways-presentations-111225.pdf
  12. ITER breeding blanket and tritium plant plans advance linked only, not reproduced
    American Nuclear Society, Nuclear Newswire · 2026
    www.ans.org/news/2026-09-25/article-8435/iter-breeding-blanket-and-tritium-plant-plans-advance/