OpenAI pauses UK data‑centre plan
OpenAI paused its proposed Stargate UK data‑centre project, citing industrial electricity prices that are far higher than in the U.S. and unresolved copyright questions tied to AI. The pause underlines that energy costs and legal uncertainty are now first‑order constraints on where AI infrastructure can be built. That combination is already shaping vendor deployment plans and may influence where companies host latency‑sensitive or regulated workloads. (thenextweb.com)
OpenAI has put its Stargate United Kingdom data-centre plan on hold after less than a year, even though the project was supposed to anchor up to 8,000 graphics processing units in northeast England in early 2026. The company told CNBC it will wait until “regulation and the cost of energy enable long-term infrastructure investment.” (cnbc.com) That project was not a sketch on a whiteboard. OpenAI announced it in September 2025 with Nvidia and Nscale, tied it to Cobalt Park in North Tyneside, and said it could expand from 8,000 to 31,000 graphics processing units over time. (cnbc.com) A data centre for artificial intelligence is basically a power station in reverse: instead of turning fuel into electricity, it turns electricity into answers, images, and software output. The more advanced the model, the more chips you stack together, and the more the power bill starts to decide the map. (tech.eu) That is where Britain ran into trouble. The International Energy Agency said electricity prices for energy-intensive industries in Europe stayed above twice United States levels in 2025, and average European wholesale power prices were roughly double the United States. (iea.org) The British government’s own energy-price series was updated on March 31, 2026 with fourth-quarter 2025 data, and industry groups have been warning that United Kingdom industrial electricity costs sit near the top of major economies. OpenAI’s pause turns that abstract complaint into a cancelled construction timetable. (data.gov.uk, soci.org) The second problem is not wires but law. The United Kingdom government says the application of current copyright law to training artificial intelligence models is “disputed,” and its March 19, 2026 consultation update says that uncertainty is already undermining investment and adoption. (gov.uk) That fight is about training data, which is the giant library of text, images, code, and audio used to teach a model patterns. If a company cannot predict whether that library will require licences, opt-outs, disclosures, or court fights, it cannot easily price a billion-pound facility built to use it. (gov.uk) The politics around that library have moved against a quick fix. CNBC reported that the United Kingdom stepped back from copyright changes that would have made it easier for artificial intelligence companies to use media content, and the government later said most consultation responses rejected its earlier broad-exception idea. (cnbc.com, gov.uk) That leaves Britain with a double handicap for artificial intelligence infrastructure. Expensive electricity raises the running cost every hour, and unresolved copyright rules raise the legal risk before the first server is switched on. (iea.org, gov.uk) OpenAI is still saying London matters, and CNBC reported that talks with Nscale have not ended. But a project once sold as part of Britain’s push for “sovereign” artificial intelligence capacity is now waiting on two older industries to settle first: power and publishing. (cnbc.com, tech.eu)
