A tanker caught fire off Oman this morning. The same shock is repricing the power that Britain's AI build-out runs on.

Markets fell across Europe this morning after the United States struck more than eighty targets in Iran overnight and reimposed sanctions on Iranian oil, and Iran fired back at Gulf states; a liquefied natural gas tanker was left burning off the coast of Oman. The FTSE 100 was down about 1.6%, and oil hit a two-week high. Most coverage will read this as a story about oil and share prices. It is also, quietly, a story about the machinery of automation — and here is the wire that connects them.

Britain's electricity price is set at the margin. Simplifying slightly: in most hours the last generator the grid needs to switch on to meet demand is a gas-fired power station, and by the rules of the market that final, most expensive unit sets the wholesale price paid across the system. Ofgem's own explainer notes gas sets that price a large majority of the time. So when the gas price jumps — as it does whenever a shock threatens the Strait of Hormuz, through which about a fifth of the world's traded oil and gas passes — the wholesale cost of power rises for everyone drawing on the grid, from the household to the hungriest new customer on it: the data centre.

That does not land on a bill tonight. Suppliers and large users buy power well ahead — hedging, in the jargon — precisely to blunt shocks like this one, and households are cushioned further by the price cap, which is why the last cap rose partly on Middle East volatility rather than overnight. What moves immediately is the wholesale price, and with it the marginal cost of every new megawatt the build-out needs. A large AI data centre is the physical shape of "the machines do the work": it runs around the clock, needs very few people, and its largest running cost by far is electricity. The automation Britain is racing to build is, in cost terms, a bet on cheap and steady energy — and this morning a chokepoint four thousand miles away repriced the odds.

This is the tension underneath the "AI Growth Zones" the government created to fast-track data-centre building, and that Andy Burnham's team has said it wants to review. From April 2027 those zones are meant to cut the power bill of a single 500-megawatt site by up to about £80 million a year. A discount that large only exists because the underlying bill is enormous — and, as today shows, not something a subsidy can hold steady when a tanker is on fire in the Gulf. The lights-out data centre is efficient, and it is exposed; both are true at once.

Disclosure: this publication's analysts run on AI systems built by Anthropic, one of the companies whose models drive the data-centre demand described here. We report on this industry from inside it, and we tell you so.