OpenAI never actually visited the place in England where the UK government said it would build a major AI computer facility. This discovery, revealed through a Freedom of Information request, calls into question a major announcement the British government made during a US state visit in September 2025.

The project was called Stargate UK. It was supposed to be a multibillion-pound AI facility involving OpenAI (an AI company), Nscale (an infrastructure company), and Nvidia (which makes computer chips). The UK government announced it on 16 September 2025 and claimed the announcement brought in a record-breaking £150 billion in investment to Britain. The facility was supposed to be built at Cobalt Park in North Tyneside, in the North East of England.

But the Freedom of Information request revealed something surprising: neither OpenAI nor Nscale ever met with local officials at the North Tyneside site. Only Nvidia visited—and that was five months after the announcement, in February 2026. Nscale said one of its executives travelled to North Tyneside, but there is no record of any actual meeting taking place.

A Guardian newspaper investigation published in July 2026 uncovered something else troubling: £20 billion of the £30 billion in investment the government had announced "appears to have been totally hypothetical." In other words, the money was promised on paper but did not actually exist in any concrete form. Someone told the Guardian that the whole thing was essentially "a government PR stunt."

What happened is straightforward: the government asked these companies to promise support for the project just in time for the announcement. The normal steps—visiting the site, talking to local authorities, arranging power contracts—were supposed to happen afterwards. They mostly did not.

Why the deal fell apart

Problems began to surface months before the July 2026 revelations. In March 2026, the Guardian reported that many UK AI deals were "phantom investments"—meaning they existed mostly on paper. That same month, OpenAI started backing away from the datacentre part of Stargate UK. By April 2026, Reuters and the BBC reported that OpenAI had officially paused the UK datacentre project. The company said the UK's expensive electricity and difficult regulatory rules made it unworkable.

That was the real issue: the conditions OpenAI needed to make the project work were not present in the UK. Building an AI datacentre requires enormous amounts of cheap electricity. The UK's electricity is expensive, and the planning process for new facilities is slow. Those problems were always going to be difficult to overcome.

Nscale faced similar issues. The Guardian had previously reported that Nscale's supercomputer project in Essex was barely started—just the empty framework of a building, no actual equipment inside. This suggests that Nscale's plans for the UK were much less advanced than the government's announcements had suggested.

There was another issue too: a March 2026 Freedom of Information request found no evidence that the UK government had even tested OpenAI's technology under a partnership signed in July 2025 that was supposed to explore using AI in courts, defence, and security. Yet ministers had said this partnership was crucial to improving public services.

The broader picture

OpenAI's Stargate UK announcement described the project as strengthening Britain's "sovereign compute capabilities." (That means having Britain's own computer power, not relying entirely on other countries.) OpenAI borrowed this language from its US Stargate programme, a $500 billion plan to build AI infrastructure across America. Since then, OpenAI has announced similar Stargate projects in the UAE—a major facility in Abu Dhabi expected to be partly operational in 2026—and in Norway, planning to install 100,000 computer chips by the end of 2026.

Those announcements came before the UK one. Both the UAE and Norway offered something the UK did not: cheaper electricity or quicker planning approval, or both. The UK's high electricity costs and slow planning system were not new problems in September 2025. They were always going to make the deal difficult.

Governments often announce big investment numbers that mix real commitments with promises that depend on future conditions. The problem happens when those two categories are blurred together. When £20 billion of a £30 billion announcement is later found to be hypothetical, and when the companies involved never even visited the site, people have a right to know. OpenAI, Nscale, and the UK government have not explained these gaps publicly. Until they do, Stargate UK looks less like a project that was paused and more like an announcement that never had a solid plan behind it.