While the industry benchmarks itself to death, Anthropic went and bought a socket. According to DigiTimes, in late September 2026 the company committed $3.4 billion to take over orders for 1.7GW of heavy-duty gas-fired generation equipment from supplier OnePWR Solutions. Not a startup bet. Not another GPU order. A place in the turbine queue.
1.7GW is not a rounding error. That is the load of a serious industrial site. And the important part: Anthropic is not building a plant from scratch — it is taking someone else's slot. Heavy gas turbine makers are booked out years ahead, and in that queue the slot costs more than the metal.
The logic is blunt. Rivals argue about chips. Anthropic closed the question everyone else meets next year: where do the kilowatts come from. DigiTimes frames it as a trend, not a one-off — "AI giants", plural. Which means the queue premium is already being paid by several players, and the queue only gets longer.
DigiTimes is a Taiwanese outlet read for supply chains, not for polished prose. Heavy turbines come from a handful of companies worldwide, and a serious slice of that market sits in Asia. Every gigawatt an AI lab locks up is a gigawatt that does not reach utilities, factories, and smaller data centers. Equipment goes where the premium is.
The first wave of the AI race was silicon. The second is electricity. A model you cannot power is a file on a disk.
And here is the uncomfortable part. Gas is not green power. Companies that publicly promise carbon neutrality are quietly buying gas generation, because wind and solar do not deliver steady baseload. That gap between the press release and the contract will echo for years — in reports, in courtrooms, and in meetings with regulators.
Prediction: The next big AI deals will read like energy contracts, not hardware purchases.
Disclaimer: supply chain leak; DigiTimes cites international media, not officially confirmed
Source: www.digitimes.com · post in Telegram