A Swedish Startup Promises Fusion Power at 2 Cents a Kilowatt-Hour. The Math Says Otherwise.
2D Fusion AB says its deuterium-deuterium reactor will sell electricity for under €0.02/kWh by 2032. Its own numbers put the capital cost alone above that price, its no-waste claim skips the tritium, and its marketing multipliers are inflated up to sixfold.
Two euro cents is the price on offer: on November 5, Swedish company 2D Fusion AB announced that its fusion power plants will produce electricity for under €0.02 per kilowatt-hour, with reactor parts shipping in the third quarter of 2029 and a first 1-gigawatt plant running by 2032.
For scale, Lazard's 2025 benchmark puts new-build nuclear at $0.141 to $0.220 per kWh, and even utility-scale solar, the cheapest power humanity has ever built, bottoms out at $0.038. A startup nobody had heard of claims it will beat solar's floor by nearly half, on a fuel cycle nobody has ever made work, with no test data attached. That is the claim, and here is what the company actually published to support it: a six-paragraph press release, a homepage, and a refusal to say anything more, with no temperature, no density, no confinement time, no Q value (the ratio of fusion power out to heating power in), no neutron yield, no peer-reviewed paper, no patent numbers, and no named customers.
Acting CEO PJ Peltola says the invention will have "replaced all fossil fuels and all nuclear fission power plants globally within 30-40 years." You can check the arithmetic yourself, because the arithmetic is the whole story.
The price fails the capital test
Start with the one number the company did give: a 1-gigawatt plant running at 90 percent capacity factor generates 7.884 terawatt-hours a year, and at €0.02 per kWh, roughly $0.023, that plant earns about $183 million a year, so now price the hardware, because capital cost is the floor under any power plant's price and it can be computed without knowing anything about 2D Fusion's plasma.
Spread a build cost over a 20-year plant life at a 5 percent discount rate, and the annual capital charge is 8.02 percent of the build cost. Run that for a range of build costs, and watch what happens to the claimed price.
| Build cost | Capital cost | Capital-only price |
|---|---|---|
| $2,000/kW, cheaper than any nuclear plant ever built | $2.0 billion | $0.0204/kWh |
| $3,000/kW, Korean APR-1400 class, cheapest nuclear on record | $3.0 billion | $0.0305/kWh |
| $5,000/kW | $5.0 billion | $0.0509/kWh |
| $10,000/kW, Hinkley Point C scale | $10.0 billion | $0.1018/kWh |
At the cheapest build cost in nuclear history, capital service alone eats 89 percent of the claimed price, before operations, maintenance, fuel processing, financing margin, or profit. At any realistic first-of-a-kind fusion build cost, capital alone runs two to four times the claimed total price. This calculation does not depend on the company's confinement scheme, its plasma temperature, or whether its reactor works at all. A gigawatt of hardware costs money, and €0.02 per kWh leaves no room for it. It does not add up.
The fuel math is real. It is also irrelevant.
One part of the pitch checks out beautifully. Deuterium-deuterium fusion averages 3.65 MeV per reaction across its two branches, which works out to about 24,456 kWh of heat per gram of deuterium, or roughly 8,071 kWh of electricity at a standard 33 percent steam-cycle conversion. A 1-gigawatt plant at 90 percent capacity factor would burn about 977 kilograms of deuterium a year. Seawater holds about 33 grams of deuterium per cubic meter, per ITER's own figures, so a year's fuel comes from roughly 29,600 cubic meters of seawater: twelve Olympic swimming pools, which is beautiful and beside the point, because fuel is the cheapest part of any nuclear plant, fission included, where uranium runs about half a cent per kWh.
A fuel cycle that costs nearly nothing does not rescue a price claim that fails on capital, and nobody ever went broke buying deuterium.
"Only helium-4" skips the tritium
Next comes the waste claim, where the release overreaches most plainly: according to the release, the process "only produces helium-4, besides pure energy," and leaves no radioactive waste. That is wrong at the level of the reaction itself.
Deuterium-deuterium fusion has two branches, and they run about equally often. One branch makes helium-3 plus a neutron, while the other makes tritium plus a proton, and tritium is radioactive, with a 12.3-year half-life. That tritium then fuses with deuterium, releasing a 14.1 MeV neutron. So a D-D plant carries a tritium inventory, a fast-neutron flux, and neutron-activated structural materials. Run the full burn cycle to completion and the net product is helium-4, which is what the company is describing. But the tritium, the neutrons, and the activated steel in between are real engineering problems that every D-T machine in history, from NIF to JET, has had to price in. "No radioactive waste" is a description of the last line of the equation, not the plant. The plant matters more.
The multipliers are inflated three to six times
2D Fusion's homepage carries three marketing multipliers, and all three can be checked with high-school physics. Burning hydrogen releases 1.48 electron volts per atom. D-D fusion releases 1.825 MeV per atom, a ratio of about 1.2 million, not the "5 million times" the site claims. That 5-million figure matches D-T fusion, which releases 8.8 MeV per atom, not the D-D cycle the company sells. Burning a molecule of octane releases about 56.7 eV, against 3.65 MeV for a D-D reaction, a ratio of about 64,000, not the claimed 200,000. And a 20-tonne truck of heavy water carries about 4,000 kilograms of deuterium, worth roughly 9.8 × 1010 kWh of heat, against about 2.85 × 105 kWh in a 30,000-litre gasoline tanker, a ratio of about 343,000, not 2 million. Each multiplier is directionally true and numerically inflated, by factors of three to six, which is the pattern that matters most: a company whose checkable numbers are all wrong in the same direction is telling you something about its uncheckable numbers, and the honest move is to listen.
The hardest fuel, with the thinnest evidence
Then there is the physics, and D-D is the hardest practical fusion fuel cycle: its cross-section runs roughly 100 times lower than D-T at reactor temperatures, per Stanford's fusion lecture notes, which means it needs a far hotter, denser, better-confined plasma for the same power. Every fusion result that has ever approached breakeven, NIF's target gain of 1.5 in December 2022 and JET's 69-megajoule shot in 2023, used D-T, and no D-D device has ever come close.
Contrast that with the real fusion news from the same week. On November 3, researchers at the DIII-D National Fusion Facility published a Physical Review Letters result demonstrating helicon current drive, a radio-wave technique for keeping tokamaks running around the clock. Tests ran at modest power in low-confinement mode, and the authors said so plainly. A modest claim, a peer-reviewed paper, and published parameters: that is what progress looks like. A company announcing commercial availability and a gigawatt plant while publishing zero parameters is selling a press release, not a reactor.
Limitations
No independent verification of 2D Fusion's claims exists, and the company declined to comment beyond its homepage. Every criticism here is arithmetic applied to the company's own published numbers, not access to its device. Capital-cost figures assume a 20-year plant life and a 5 percent discount rate; a 40-year life at 3 percent would cut the capital-only price by roughly a third, still above the claimed price at any build cost over about $3,000 per kW. The 33 percent conversion figure is a standard steam-cycle assumption, and a direct-conversion scheme could improve it, though gigawatt-scale direct conversion is itself unproven. Multiplier checks depend on the comparison basis, stated explicitly for each. Lazard's figures are U.S. new-build estimates from June 2025; European costs differ, but the order of magnitude stands.
Strongest case against
Fairness demands the steelman, because it is a good one. Fusion history is full of outsiders dismissed too early, and 2D Fusion describes its approach as "low-temperature" fusion, a regime the standard tokamak numbers above do not describe. If its confinement scheme is genuinely novel, then D-T cross-section comparisons and tokamak capital costs are the wrong benchmarks, and the absence of published parameters could reflect trade secrecy rather than emptiness. Trade secrecy is compatible with a working device. But commercial availability is a public claim that invites public evidence, and selling reactor parts for 2029 delivery while publishing zero test data, zero customers, and zero physics is not how any fusion company that later delivered, CFS, Helion, TAE, or General Fusion, behaved at the equivalent stage. And the price claim fails on arithmetic that does not depend on the confinement scheme at all. Whatever the plasma does, a gigawatt of hardware costs money.
What you can do
If you follow fusion, ask for four numbers before believing any announcement, temperature, density, confinement time, and Q, and treat any company that cannot produce all four as a press release with a logo rather than a reactor with a future. If you evaluate an energy claim, compare the price against Lazard's levelized-cost table first: a number below solar's $0.038 per kWh floor needs extraordinary evidence. If you invest in energy, mark the calendar for the third quarter of 2029, 2D Fusion's own deadline for first deliveries. A missed delivery date is the cheapest falsification test in the industry, and it costs nothing to wait for it.
Bottom line
A Swedish startup says it will sell fusion electricity for under two euro cents a kilowatt-hour by 2032, from a fuel cycle nobody has ever made work, with no published test data, and its fuel math is real: a gigawatt plant would burn under a tonne of deuterium a year, drawn from twelve swimming pools' worth of seawater. Everything else fails on its own numbers: capital service alone exceeds the claimed price at any build cost above the cheapest nuclear plant ever built, the "no radioactive waste" claim skips the tritium and the neutrons, and the marketing multipliers run three to six times high. Two cents is a lovely price, and the company has not shown a single number that makes it possible.