General Fusion Hit 1 keV. The Hard Part Is the Next Factor of Ten

On 8 October, General Fusion pushed a deuterium plasma to 1.18 keV, roughly 12.6 million °C, by crushing it inside a solid lithium liner. The machine is Lawson Machine 26: a 1.76-metre liner imploded by theta-pinch coils at the company's Richmond, British Columbia site.

UKAEA measured the same plasma independently with Thomson scattering and recorded 1.09 keV. That is the technique British physicists used in 1969 to confirm 1 keV electrons in a Soviet tokamak.

The temperature is 58 years old. The method is not. General Fusion compresses for milliseconds behind a metal wall, without superconducting magnets and without high-power lasers. On capital cost, that should be the cheaper route to fusion. On evidence, it is the least proven approach now claiming the benchmark.

Both sides of that ledger arrived in the same week: a physics number measured by a national laboratory, and a company asking investors to wait until the 2030s for a plant. Which one carries more weight decides whether 1 keV is a turning point or a press release with error bars.

The case that the number is real

Fusion announcements usually fail at the measurement, so start there. Three diagnostics agree on this shot. Filtered X-ray diodes put peak electron temperature at 1,180 ± 65 eV. Thomson scattering, the gold standard, recorded 1,090 ± 40 eV at a radial compression of 2.35. An absolute extreme ultraviolet ratio extended the series to 2.75 compression, where the company reports 1.2 keV.

Ion temperature came out at 459 ± 33 eV, inferred from neutron yield and supported by carbon line Doppler spectroscopy. Seventeen Mirnov probes constrained the equilibrium reconstruction. None of this is a single-instrument claim, which is the usual weak point in private fusion results.

The shot, LMC-12, is number 8131 in the LM26 campaign. Radial compression of 2.75× raised electron temperature more than fivefold and ion temperature up to twofold, with density and magnetic field rising as a stable plasma should. Compressional heating delivered 9.6 MW to electrons and 3.0 MW to ions, on top of 8.1 MW of Ohmic heating. In the previous best shot the comparable figures were 4.3 MW and 2.4 MW.

Then there is the engineering argument, which is the actual product. Magnetized target fusion sidesteps two of the most expensive subsystems in the fusion bill of materials. No superconducting magnets means no cryoplant and no dependence on high-temperature superconductor supply. No lasers means no megajoule-scale optics to damage and replace. As we wrote in October, energy hardware economics now turns on substituting ordinary materials for exotic ones, and this is the same trade in a different machine.

The company's plant design replaces the solid liner with a liquid lithium wall. That wall absorbs neutrons, breeds fuel and carries heat to a turbine, which turns three of fusion's hardest components into one circulating fluid. The physics result matters because it keeps that architecture alive. The architecture matters because it is the only part with a plausible cost story.

General Fusion has run this argument for two decades: more than 200,000 plasma experiments and 210 issued or pending patents, according to its investor material. It listed in July 2026 with about $150 million in cash to fund LM26 through the 2028 Lawson campaign. On the day of the announcement the shares rose 12.96% to $8.58, according to Benzinga.

What 1 keV leaves untouched

Ten keV is the number on the wall, and the distance is not a rounding error. Electron temperature has to rise by a factor of ten, to roughly 100 million °C. The Lawson criterion then multiplies temperature by density and confinement time, so a single axis improving tells you less than it seems.

The ion number is the sharper constraint. At 459 eV, ions sit well below the electrons that got the headline. Fusion rate is governed by the ion population, and the gap between electron and ion temperature is where pulsed concepts usually lose their advantage.

One shot is also one shot. LMC-12 is the best result of a campaign, not a repeatable operating point. A plant needs thousands of pulses at that performance, and the variance between them is unpublished.

Only a few fusion companies have surpassed the industry-recognised milestone of 1 keV, all using other approaches.— UKAEA and General Fusion joint release, 8 October 2026

Then the consumable problem. What happens to a liner after each pulse? In the company's plant design it is consumed and replaced. That makes the liner a recurring cost rather than a capital asset, and it turns the economics into a question about repetition rate, liner manufacturing tolerance and lithium inventory.

Fuel is the second unproven loop. The experiment runs on deuterium. A power plant needs tritium, which the liquid wall is meant to breed from neutron capture. A closed breeding cycle has not been demonstrated at power-plant scale, and it is a licensing question as much as an engineering one.

Both papers describing this result are submitted for peer review, not published. The company's own release carries a cautionary note: the results are preliminary and do not guarantee the 10 keV target or the Lawson criterion. Read the milestone as a verified measurement with an unverified destination.

Capital is the quiet constraint. About $150 million funds a demonstration programme through 2028. A first-of-a-kind plant targeted at 2035 is a different order of spending, and the company has no product revenue between now and then.

Why the money moved before the physics

General Fusion is not raising against a quiet field. Type One Energy closed a $200 million Series B in early October, according to The Fusion Report. Germany committed a €2.5 billion programme on 7 October to pull private capital into fusion, a week after the LM26 result. Stellarators and tokamaks absorb that money into magnets and shielding. Pulsed concepts absorb it into capacitors, liners and shot time.

That difference in capital intensity is the real competitive claim inside the 1 keV announcement. A $150 million budget buys a meaningful demonstration programme in a machine built in under two years. The same sum would be a rounding error on a burning-plasma tokamak.

Public markets add a second variable. General Fusion describes itself as the first listed pure-play fusion company, so its share price now doubles as a daily referendum on the sector's timeline. A 13% move on a physics result is a small sample of that behaviour, and it cuts both ways when a milestone slips.

None of this changes the physics. It does explain why a 1 keV electron temperature, first recorded in 1968, is worth a press release in 2026: the cost of finding out whether the cheap route works has fallen far enough for private balance sheets to fund the experiment.

Where the two readings meet

The 1968 comparison is the useful frame, and not because it diminishes the result. Soviet physicists reported 1 keV in the T3 tokamak; a British team confirmed it with Thomson scattering a year later. Fifty-eight years on, the benchmark has moved from national laboratories to private machines. What changed is who pays and how fast they iterate.

That is also the honest limit of the news. The interesting variable in 2026 is capital cost per keV, and neither tokamaks nor magnetized target fusion has a published cost curve yet. The metal-wall route bets that consumables are cheaper than magnets at scale. The bet is now testable, which is more than it was a month ago.

The commercial context has shifted underneath the physics. Data centre operators are signing firm capacity for the 2030s, and the queue they are joining includes gas, batteries, small modular reactors and geothermal. Fusion enters that queue with a 2035 plant date and no operating revenue.

A verified 1 keV result buys credibility in that queue. It does not shorten it.

The electrons are hot. The schedule is still a bet.