WEST tokamak record: 22-minute plasma shifts fusion focus to tungsten

A record that is about endurance

On February 12, 2025, the WEST tokamak at the CEA's Cadarache site in southern France held a hydrogen plasma for 1,337 seconds, or about 22 minutes. That set a new world record for how long a tokamak has sustained a plasma.1213 The pulse ran 25% longer than the 1,066-second mark China's EAST tokamak had set only weeks before, in January 2025.1113 More than a year and a half later, the result still comes up often. EUROfusion's 2025 review listed it among the consortium's main achievements, and outlets were republishing the story as recently as May 2026.71

Popular coverage presents the result as a fusion breakthrough. A better description is a materials and control milestone that happened inside a fusion machine. The plasma reached about 50 million degrees Celsius, roughly three times the temperature of the Sun's core.11 Commercial fusion conditions are usually put at 100 to 150 million degrees.1 The plasma was also not producing net energy. A lower hybrid radiofrequency antenna pumped 2 MW of heating power into it for the whole pulse.12 What the shot showed is that a tungsten-walled machine can stay stable for a long time and that its plasma-facing parts can take the heat. Any future power plant has to clear that hurdle, but the hurdle is about engineering, not about getting energy out.

What actually happened during shot 61299

The CEA's technical account has more detail than most news stories. The team scheduled two days, February 12 and 13, specifically to try to pass 1,000 seconds, which was one of WEST's main goals.12 An earlier hydrogen pulse that day lasted 819 seconds. The next one, plasma number 61299, ran to 1,337 seconds, and about 2.6 gigajoules of energy was injected by the heating system and removed by the actively cooled tungsten components.12 Throughout the pulse, the control system adjusted heating power in real time to hold the plasma current at 215 kA. Piezoelectric microvalves injected hydrogen to keep the density steady.12

One detail barely made it into the coverage. About 1,040 seconds in, one of the seven klystrons powering the heating antenna cut out. The plasma settled into a different regime and carried on until the lower hybrid system shut down at second 1,337.12 The CEA said it was still analysing what happened.12 So the record pulse included an equipment failure partway through. In practical terms, how the plasma rode through that failure may matter more than the final number. Long pulses depend on control systems that can react to unexpected events, including sudden overheating of wall components, and the CEA said as much.12 The next day, a deuterium plasma lasted 921 seconds, which suggests the scenario can be repeated.12

Where the coverage diverges

The reports agree on the main facts: 1,337 seconds, 50 million degrees, 2 MW of heating, and a 25% gain over EAST.111317 They differ on framing, and some of the differences matter.

Some outlets described the event as a "plasma reaction" or "sustained fusion," which suggests fusion reactions were the point of the test.17 The CEA's own wording is narrower. It describes sustaining a hydrogen plasma and stresses plasma control and the durability of plasma-facing components.1217 Readers in the New Atlas comments section noticed the gap and pointed out that a plasma kept going with outside heating is not self-sustaining.1 That reading is correct.

Accounts of the EAST record it replaced also conflict. The ITER Organization says EAST held its plasma at a temperature of "nearly 70 million °C."18 A later article from mid-2026 says EAST passed 100 million degrees during the same run.6 These figures probably refer to different plasma species or measurement points, but the inconsistency should make readers wary of using temperature claims to rank records. The ITER account also notes that EAST ran in high-confinement mode with a tungsten divertor and lithium injection.18 WEST's record was set under different conditions, so calling it a straightforward head-to-head contest oversimplifies things.

The best way to read the competition comes from a fusion forum: there are many separate records, held by different machines, much like events at the Olympics.9 WEST holds duration. JET holds the record for total fusion energy produced. The National Ignition Facility holds the record for energy gain.9 Germany's Wendelstein 7-X stellarator set a world record in 2025 for the triple product over a 43-second high-performance discharge.7 A power plant would need to achieve all of these at once.

The materials science story underneath

The "W" in WEST is the chemical symbol for tungsten, and tungsten is the main reason the machine exists. Tore Supra, the machine WEST replaced, was rebuilt from 2013 with all-metal cladding, mainly to test tungsten divertor technology for ITER.26 ITER decided in 2023 to switch its first wall from beryllium to tungsten, which makes WEST's data directly useful for ITER.2

Tungsten is hard to work with. Even small amounts of tungsten that erode from the wall into the plasma can cool it quickly or shut it down altogether.202 A Princeton Plasma Physics Laboratory researcher compared running a tungsten-wall machine with running a carbon one to petting a wild lion instead of a house cat.2 That is why diagnostics have become central. PPPL developed a multi-energy soft X-ray camera that can track tungsten impurities from wall to core.2 In 2026 it also began installing two additional X-ray crystal spectroscopy systems on WEST to map temperature, rotation and tungsten levels across the plasma.

The most practical materials results so far appeared in a June 2026 Nuclear Fusion paper. It reports that nearly 16,000 ITER-grade tungsten blocks lining WEST's lower divertor received about 18 hours of plasma exposure in total, with peak heat fluxes up to 12 MW per square metre.23 The authors also describe the first controlled tests of runaway-electron impacts on tungsten in WEST, which caused major damage even though tungsten has a very high melting point.23 That result should temper the more upbeat coverage. The 22-minute record shows the wall can survive a quiet pulse. Surviving disruptions is still an open question. The CEA plans to raise heating power from 2 MW to 10 MW over pulses of around 1,000 seconds, which it says corresponds to gigawatt-scale fusion power in a machine the size of ITER. The stated aim is to measure how long the tungsten components last.12

AI enters the control room

The research around WEST also shows the growing role of AI in science. A transformer-based model trained on 550 WEST discharges and published in Nuclear Fusion in March 2026 predicts six key global plasma parameters. It uses only inputs that operators set before a shot, such as coil currents and heating power.2224 It reached an average R² of 0.94 and returns predictions in about a tenth of a second. The authors say this could speed up discharge planning and eventually support real-time control.22 They are open about its limits. It predicts the safety-factor quantities less accurately, and it was trained only on stable discharges with good wall conditions.25

Similar work is under way in the wider field. Researchers are building machine-learning tools that spot the early signs of tearing-mode instabilities before standard diagnostics can see them. This work feeds into designing the trigger for ITER's disruption mitigation system.27 Other reports describe reinforcement-learning controllers that handle heating systems and predict edge energy bursts.28 The link to the record is clear. The CEA's account of shot 61299 involves control systems adjusting on the fly and a plasma reorganising itself after a hardware failure.12 Hours-long pulses, which are the CEA's next goal, will probably depend heavily on this kind of predictive control.11

Some commentary also presents the record as a mathematical breakthrough. That claim has no support. Nothing in the reporting suggests a new theoretical proof. The progress comes from experiments, engineering and statistical models, and the data-driven models are attractive precisely because full physics simulations cost too much computing time for fast scenario design.24

Why it matters now

The pressure on fusion has grown since the shot. Private fusion companies raised a record $4.48 billion in the year to July 2026, and AI data centres are now being pitched as their first customers.10 China has completed a 582-ton superconducting magnet for its BEST tokamak.10 Helion says its Polaris prototype reached 150 million degrees with deuterium-tritium fuel.8 Against these headline numbers, a 50-million-degree plasma driven by outside heating can look modest.

The WEST record still gives ITER something it needs. ITER's current schedule calls for deuterium-tritium operations in 2039.6 Before then, it needs evidence that tungsten walls, real-time control and long-pulse heating can work together for many minutes at a time. WEST is the closest thing ITER has to a test bed for that combination. The 1,337-second figure is the part people remember, but the lasting value lies in the cracking studies, the impurity measurements and the machine-learning models built from WEST's discharges.2322