Anthropic secures 1.7 GW of gas generators in a 3.4 billion dollar deal

📅 November 9, 2026 • Financial Times Anthropic has agreed to pay 3.4 billion dollars for gas-fired generators to power its data centers, the Financial Times reported on November 8. Reuters and Bloomberg carried the story the same day, and TechCrunch added further detail on November 9. The FT described the transaction as the largest known single purchase of power-generation equipment ever made by an AI lab. The deal covers commitments for 1.7 GW of heavy-duty generators, secured in late September from OnePWR Solutions, an equipment supplier based in San Antonio, Texas. The machines are built by Caterpillar, and the agreement includes engineering and construction work for installation. Per the FT, Anthropic did not acquire OnePWR, and none of the supplier employees moved to Anthropic. TechCrunch reports that Anthropic first planned to buy the units from OnePWR, then went directly to Caterpillar and took over the existing orders. The purchase marks a shift for a company that has historically rented compute rather than owning physical infrastructure, and it now places power equipment among the assets Anthropic holds directly. The context helps explain the move. Per the FT and TechCrunch, backlogs for new gas turbines stretch into the early 2030s, and grid connection queues can be just as long. Earlier this year, per the FT, xAI bought mobile gas turbines worth about 2 billion dollars, and Alphabet acquired energy developer Intersect in late 2025. The supplier side shows the same demand. Per Caterpillar's third-quarter release, filed October 29, Power Generation sales rose 30 percent to 3.421 billion dollars in Q3, driven mainly by turbines and large engines primarily in data center applications. CEO Joe Creed described a record 78 billion dollar backlog. For engineering and operations teams, the deal signals that capacity planning now starts at the power layer. Building AI infrastructure increasingly means treating generation equipment, delivery timelines and installation engineering as one procurement cycle, because waiting for a grid connection can take as long as waiting for the turbines themselves. Teams planning AI buildouts now need to factor equipment lead times into roadmaps the way they once factored GPU availability. Sources: TechCrunch, CNBC, Bloomberg, Financial Times, SEC EDGAR, The Company Chronicle Continue the conversation: https://t.me/AIAutomationArchitecture · Bluesky