Anthropic, the AI company planning a public offering this year, is purchasing $3.4 billion worth of natural gas turbines. The Caterpillar generators involved in the purchase can collectively produce up to 1.7 gigawatts of electricity.
Shift from renting to owning power capacity
Historically Anthropic has preferred renting infrastructure. In its S‑1 filing in May the company disclosed it is paying SpaceX subsidiary xAI $1.25 billion per month to use xAI’s computing capacity, a payment schedule that runs through May 2029. The filing also stated Anthropic expects to enter into additional similar services contracts, indicating continued reliance on rented compute.
Despite that approach, rapidly rising demand for compute has increased uncertainty in the market for power. Lead times for natural gas turbines extend into the early 2030s, and waiting lists to connect generation to the electrical grid can be similarly long. Those constraints appear to have pushed Anthropic toward a direct purchase.
How the deal came about
According to the Financial Times, the turbines were originally ordered a couple of years ago by equipment supplier OnePWR, but OnePWR could not obtain financing to complete the purchase. Anthropic had planned to buy the equipment from OnePWR, but instead negotiated directly with the manufacturer, Caterpillar.
The report does not explain why OnePWR failed to secure financing despite Anthropic’s apparent interest as a customer. Anthropic’s direct purchase suggests the company preferred to retain the generation capacity itself rather than risk a competitor acquiring it.
Industry context and risks
Other major cloud and tech companies have also been increasing on‑site natural gas capacity. Google, Amazon, Microsoft, and Meta have all announced plans to build large natural gas power plants to run their data centers rather than relying solely on the public grid.
This trend has helped revive fracking and, if buildouts continue as projected, U.S. data centers alone could consume more natural gas by 2035 than Germany and Japan combined. Greater reliance on natural gas also exposes hyperscalers to price volatility: one analyst cited in the report expects prices in some markets could more than triple by the end of the decade. If that materializes, Anthropic could face significantly higher operating costs after its large upfront purchase.
Conclusion
Anthropic’s $3.4 billion turbine purchase highlights how securing reliable power has become a strategic priority for companies scaling large AI models. The move reflects supply constraints in generation equipment and grid connections, and it underscores broader industry shifts and the economic risks tied to heavy dependence on natural gas.
