The Iran war is squeezing AI's power supply

THE LARGEST SUPPLY disruption in the history of the global oil market, as the International Energy Agency calls it, has created an unlikely new casualty: the artificial intelligence boom. Since the U.S.-Israeli strikes on Iran began on February 28, Brent crude has surged past $100 a barrel, the Strait of Hormuz has been effectively closed to commercial shipping, and Iranian drone attacks have crippled Qatar's Ras Laffan facility, the world's largest LNG export complex. The IEA's director, Fatih Birol, assessed the crisis as worse than the two 1970s oil shocks and the 2022 gas crunch "put together." For the tech industry, the timing could scarcely be worse. America's hyperscalers were already facing a political revolt over data center energy consumption when a war on the other side of the world made their electricity problem dramatically more expensive and their chip supply chain dramatically more fragile.

The collision is three-dimensional. First, power costs. The U.S. had nearly tripled its gas-fired power capacity in development in 2025, reaching 252 gigawatts, according to the Global Energy Monitor, with the vast majority earmarked for data center load. Natural gas is the marginal fuel for American electricity generation; when gas prices rise, so do power bills. Even before the war, U.S. residential electricity prices had climbed more than 36% since 2020, to 17.44 cents per kilowatt-hour, with the EIA projecting further increases. The Iran conflict has now added an estimated $14 to $18 per barrel risk premium to crude, per Goldman Sachs, while European gas prices have doubled since March 1. U.S. natural gas prices have been more insulated (they reflect domestic supply and demand rather than global LNG markets), but the long-term trajectory is unambiguous: LNG exports are forecast to rise 50% by 2027 compared to 2024, tightening the domestic market just as data centers are layering on unprecedented new load. AI facilities consume three to five times more electricity than conventional data centers. Every dollar added to fuel costs ripples directly into the total cost of ownership for hyperscale operators.