Atlas Energy Solutions surges as much as 19% after landing $613M in contracts with frontier AI lab

Atlas Energy Solutions just proved that the fastest way to get investors excited in 2026 isn’t building AI models. It’s powering them.

The company announced two cost reimbursement agreements with an unnamed frontier AI lab totaling approximately $613.5 million, sending shares surging as much as 19% intraday on September 25. The deals cover balance-of-plant infrastructure and 283 MW of Caterpillar natural gas generators, with an additional 328 MW commitment under an existing Global Framework Agreement with Caterpillar slated for delivery in 2027.

From sand to servers

Atlas Energy Solutions (NYSE: AESI) started life as a proppant supplier for oil and gas fracking operations in the Permian Basin. The company has been building out its behind-the-meter power business, a model where electricity is generated on-site rather than pulled from the traditional grid.

CEO John Turner framed the agreements as stepping stones toward long-term power purchase agreements with the AI lab. The cost reimbursement structure means the AI company fronts the capital for equipment purchases, which de-risks Atlas’s supply chain and eases financing pressure.

The math behind the megawatts

The combined capacity from these deals reaches 611 MW when you add the 283 MW of new Caterpillar generators to the 328 MW already committed under the existing GFA. The stock traded to approximately $13.13 at its intraday peak, reflecting gains between 13% and 19% depending on the reference point.

This isn’t Atlas’s first foray into the space. The company secured a 120 MW power purchase agreement earlier in 2026, and its broader partnership with Caterpillar targets multi-gigawatt capacity expansion by 2027.

For investors watching the energy-AI convergence, the key metric to track isn’t just megawatts contracted. It’s the conversion rate from cost reimbursement agreements to long-term PPAs. Turner’s comments suggest Atlas views these agreements as a pathway to PPAs, but until those longer-term contracts are signed, the revenue visibility remains provisional.

The risk is straightforward: if the AI lab delays or cancels its data center buildout, Atlas could find itself holding generating capacity without a buyer. The cost reimbursement structure mitigates equipment risk, but the broader market risk of AI spending pullbacks remains a factor that no contract structure can fully eliminate.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.