The Frothiest AI Trade Might Be in… Energy, Not Tech

With input from the Wall Street Journal and BBC.

Forget stretched tech multiples — the frothiest AI bet right now is in energy, where concept stocks with no revenue are being priced like tomorrow’s utilities. A clutch of pre-revenue power players has swelled past $45 billion in market value on the promise they’ll one day feed AI-hungry data centers.

The standout is Oklo, the Sam Altman–backed nuclear startup whose shares have jumped roughly eightfold this year, putting its market cap near $26 billion — making it the largest US-listed company with zero trailing revenue. Oklo is developing sodium-cooled small modular reactors using enriched uranium that’s tough to source, has no NRC license, no binding power contracts, and isn’t expected to see substantial revenue until 2028. Fermi arrived with a near–$19 billion debut and still commands north of $17 billion despite having no binding customers. It wants to build 11 gigawatts of data-center power — about New Mexico’s capacity — via gas, nuclear, solar and batteries, but equipment secured so far covers only about five percent of that goal.

Smaller nuclear dreams are catching bids, too. Nano Nuclear Energy has more than doubled this year and is valued above $2 billion; Terra Innovatum, fresh from a SPAC listing, tops $1 billion. Even companies with revenue are riding the AI tide without profits on the horizon: NuScale has surged about 155% and Plug Power roughly 90% year to date, yet Wall Street doesn’t expect either to turn profitable before 2030.

Profit-makers look expensive as well. Bloom Energy has rocketed more than 400% and trades around 133 times forward earnings after Brookfield said it would invest up to $5 billion to deploy Bloom’s tech. Centrus Energy sits near 99 times forward earnings. The logic behind all this is simple: AI’s power appetite is exploding, and nuclear offers 24/7, carbon-free juice. The reality is messier. Most SMR designs remain on paper, licensing is slow, fuel supply is constrained, and public acceptance is mixed. Three Mile Island still looms in the American imagination, even though it caused no fatalities, and local backlash is real — one New York town recently banned nuclear generation after a crypto-mining reactor proposal. Waste remains a thorn: Stanford researchers have argued SMRs could generate more radioactive waste per unit of power than conventional reactors.

Tech and energy aren’t waiting. Microsoft signed a deal tied to power from the remaining Three Mile Island reactor and joined the World Nuclear Association. Google partnered with Kairos Power in Oak Ridge, Tennessee, aiming for 50 megawatts by 2030, but even a tenfold increase by 2035 won’t dent near-term AI demand. The International Energy Agency says data-center electricity use could roughly double within five years. Experts caution that SMRs are years from being broadly economical, and they won’t solve next year’s crunch.

If AI enthusiasm fades, the longest drop awaits companies with no revenue and no contracts. Even if the boom persists, there’s unlikely to be a single silver bullet. As Microsoft put it in a recent brief, no one technology will meet the world’s electricity and decarbonization needs. For investors piling into pre-revenue power plays, that’s both the dream — and the risk.