Crusoe Is Priced 670% Higher in One Year — Because the AI Race Is Now About Power, Not Chips

The numbers are brutal in their simplicity: according to private-market data, Crusoe Energy Systems' valuation has risen roughly 670% in a year — nearly a sevenfold increase — even as the company builds out some of the most sought-after AI data centers in the United States. A firm that started by capturing gas from oil wells in Texas and North Dakota now trades, at least on paper, as something other than an oilfield-services company. It looks like the pricing of a utility with predictable revenue streams.

That is precisely the point. Crusoe calls itself "the AI factory company," positioning itself first as an energy business and only second as a compute provider. And in an industry where everyone talks about chip scarcity, it is a reminder that the real bottleneck has moved: the process of getting new power connected to the grid now takes longer than getting hold of GPUs.

The Bottleneck Is No Longer the Chip

The logic behind the repricing is easy to sketch. A GPU order can be accelerated with money and priority. A power line cannot. Interconnection queues at American utilities stretch across years, build times for new transmission and generation are measured in years, and the best sites — with access to power, water, and fiber — are already taken.

This is where Crusoe inverts the stack. Instead of finding a site and hoping for electricity, the company secures the energy source first — historically gas that would otherwise have been flared — and then builds computing power around it. It is this energy-first model, combined with large-scale GPU rental, that investors are now pricing in. The 670% figure is a broker mark in an illiquid private market, not an official valuation, and should be read as a signal of demand rather than confirmed value. But the signal is clear enough: capital is betting that whoever controls the electrons sets the tempo for the entire AI industry.

The Reactor Meant to Meet the GPUs

The most concrete evidence of the ambition arrives soon from an unexpected place. A sodium-cooled nuclear microreactor has reportedly reached "criticality" — a self-sustaining chain reaction — and is now set to be tested against a real workload at a Crusoe AI data center, with GPUs powered directly by the reactor. The point is not that microreactors will replace the grid. The point is to test whether AI factories can move to where the power is, instead of waiting for power to arrive where the factory stands.

The same logic runs through the rest of the industry. Lancium, the company behind the Abilene campus in Texas that anchors part of the Stargate buildout, has partnered with NVIDIA on gigawatt-scale AI factories — a gigawatt is on the order of a small power plant. So the pattern is not unique to Crusoe: access to energy has become the core strategy for anyone building AI at scale.

It also explains the capital raising in adjacent industries. Form Energy, which is developing iron-air batteries designed to deliver power continuously for 100 hours, has reportedly raised more than $2 billion. Long-duration storage for AI loads is no longer a climate project — it is infrastructure for computing.

The Lawsuit That Shows Who Pays

But here is where it breaks. In Nevada, the utility NV Energy has filed a lawsuit of a new kind against data center developer Tract, accusing the company of trying to push the costs of grid expansion onto consumers. The case is the first of its kind, and it poses a question that will follow the AI construction boom everywhere: if AI factories require a bigger and more expensive power grid, should households and small businesses pay for it?

This is the political counterforce that Crusoe's price surge has so far been able to ignore. The strategy of standing off the grid — microreactors, wellhead gas, self-generation — is partly an answer to interconnection queues, but also an answer to this politics: perhaps you can build your way around both the queue and the cost dispute. But self-generation carries its own risk. A reactor meant to keep GPUs running is a development project, not a finished product, and none of it has been proven at scale.

The Advantage Holds Only While Power Is Scarce

There are two ways to read Crusoe's price. The optimistic one: the company owns the gateway to the scarcest resource in the AI economy, and its valuation reflects that. The skeptical one: the price reflects a scarcity that could evaporate — because the grid catches up on its backlog, because reactors and batteries make power portable, or because politicians, as in Nevada, step in to decide who bears the costs.

Whichever reading proves correct, the shift has already happened. The AI race is still described as a race for computing power. In practice, it is a race for megawatts — and the next major reckoning over AI will happen not inside a model, but on the power grid.