Anthropic buys $3.4 billion in gas turbines

The AI company is shifting beyond rented compute and locking in up to 1.7 gigawatts of power as turbine supply and grid access tighten.

Anthropic is spending $3.4 billion on natural gas turbines, a sign of how sharply the AI industry’s power needs have grown. The company, which has mostly preferred to rent computing capacity rather than own infrastructure, is now buying equipment that can generate up to 1.7 gigawatts of electricity. The scale is striking on its own, but the deal also shows how power shortages are starting to shape AI business decisions.

It also reflects a broader tension in the AI boom. Companies need far more electricity than the existing system can easily provide, and waits for both turbines and grid connections have become long and uncertain. In Anthropic’s case, according to IT-PUB News, the purchase appears to be a fallback after the market for natural gas equipment became too constrained to ignore.

Anthropic shifts from rented compute to direct power capacity

Anthropic is planning to go public this year, and the turbine purchase suggests it is preparing for a more infrastructure-heavy future. According to the Financial Times report cited in the source, Anthropic is buying Caterpillar generators that equipment supplier OnePWR had originally ordered.

That marks a notable change for a company that has generally avoided buying physical infrastructure. In May, Anthropic disclosed in its S-1 filing that it is paying xAI, a SpaceX subsidiary, $1.25 billion per month through May 2029 to use xAI’s computing power.

The filing also said Anthropic expects to enter into additional similar services contracts. In other words, the company has so far been willing to rely on rented computing power and sign more deals like that when needed.

Tight turbine supply and grid delays raise the stakes

The reason the purchase drew attention is straightforward: AI companies are running into a power crunch. As demand for compute rises, the supply of energy — and the equipment needed to generate it — has become harder to secure.

The source says the backlog for natural gas turbines now stretches into the early 2030s, and the queues to connect to the electrical grid can be just as long. That turns access to power into a strategic issue, not just an operational one. For AI firms, the ability to train and run models depends not only on chips and data centers, but also on whether they can actually get electricity.

Anthropic’s move looks like a response to that pressure. The source describes it as a “last resort,” suggesting the company may have decided that renting alone was no longer enough to guarantee the capacity it needs.

OnePWR’s financing trouble reshaped the purchase

The turbines Anthropic is buying were originally ordered by OnePWR a couple of years ago, but the supplier reportedly could not find financing to complete the purchase. Anthropic had planned to buy the turbines from OnePWR, but instead went directly to Caterpillar, the manufacturer.

The source does not explain why OnePWR was unable to secure financing despite Anthropic’s apparent interest in the equipment. Still, the sequence suggests Anthropic may have been trying to lock in capacity before someone else did.

The report says Anthropic likely wanted to keep the generating capacity to itself rather than leave room for a competitor to buy it. That detail shows how scarce this equipment has become. In the current market, turbines are not just industrial hardware; they are a way to secure future computing power.

More AI companies are leaning on natural gas

Anthropic is not alone in turning to natural gas to support AI growth. The source says Google, Amazon, Microsoft, and Meta have all announced plans to build large natural gas power plants to support their data centers on site instead of drawing power from the public grid.

That trend is already having wider consequences. The source says the buying spree has given a second wind to fracking, the drilling method used to extract natural gas from underground rock. It also points to a possible long-term strain on energy markets: if the planned buildout continues, U.S. data centers alone could consume more natural gas than Germany and Japan combined by 2035.

There is a financial risk as well. Depending on how the market develops, hyperscalers and AI companies could face much higher fuel costs. One analyst expects natural gas prices could more than triple by the end of the decade in some markets. If that happens, Anthropic’s expensive turbine bet could become even more costly.

For now, the deal shows how far AI companies are willing to go to secure the electricity behind their products. What started as a race for chips and cloud capacity is also becoming a race for fuel, generators, and grid access.