OpenAI never visited one of Stargate UK's designated sites, a Freedom of Information request has revealed, casting serious doubt on the substance behind a project the UK government used to anchor its headline AI investment narrative during Donald Trump's September 2025 state visit to London.

The Stargate UK project — a multibillion-pound datacentre initiative involving OpenAI, Nscale, and Nvidia — was announced on 16 September 2025 alongside a broader package the government claimed represented a record-breaking £150 billion in total investment unlocked by the visit. The North East of England was positioned as the project's anchor geography, with Cobalt Park in North Tyneside designated an "AI growth zone." A dedicated North East AI Growth Zone taskforce followed within days.

What the FoI request now shows is that neither OpenAI nor Nscale ever met with local authorities at the North Tyneside site. Only Nvidia visited the North East combined authority — and that visit happened in February 2026, five months after the announcement. Nscale's spokesperson said the company's chief commercial officer did travel to North Tyneside but could not confirm whether any meetings took place; no record of one exists.

The Guardian investigation published on 4 July 2026 also reports that £20 billion of the £30 billion in Stargate UK investment touted by the government "appears to have been totally hypothetical." A source told the Guardian that Nscale "were pretty much told to back the Stargate project" and that it "was effectively just a government PR stunt."

The timeline of unravelling

The cracks appeared well before this week's reporting. In March 2026, a separate Guardian investigation found that many UK AI deals were "phantom investments." That same month, reporting indicated OpenAI appeared to be stepping back from the datacentre expansion component of the Stargate UK deal. By April 2026, Reuters and the BBC both reported that OpenAI had formally paused its UK data centre project, citing an unfavourable regulatory environment and high energy costs.

Those April reports confirmed what the March coverage had signalled: the commercial conditions OpenAI needed were not in place. The UK's electricity costs for large-scale compute — already a known constraint for hyperscale operators — combined with planning and regulatory friction appear to have made the economics unworkable, at least on OpenAI's timeline.

The pattern extends to Nscale independently. The Guardian had previously reported that Nscale's supercomputer project in Essex remained "a scaffolding yard." That detail, read alongside the FoI findings on North Tyneside, suggests Nscale's infrastructure ambitions in the UK were considerably less advanced than the government's announcements implied.

A parallel FoI finding, reported in March 2026, found no evidence the UK government had trialled any OpenAI technology under the partnership signed in July 2025 to explore AI applications across justice, defence, and security — despite ministers describing the deal as central to public service reform.

What the Stargate brand actually meant

OpenAI's Stargate UK blog post framed the partnership as strengthening the UK's "sovereign compute capabilities." The framing borrowed directly from the US Stargate programme, under which OpenAI committed $500 billion in domestic AI infrastructure investment. The brand has since been extended to the UAE — a 1GW cluster in Abu Dhabi with 200MW expected live in 2026 — and to Norway, targeting 100,000 Nvidia GPUs by end of 2026.

Those projects were announced in May and July 2025, respectively. Both are in markets that have either more favourable energy pricing, streamlined permitting, or both. The UK's position in that competitive landscape — high grid costs, a planning system under reform, and a regulatory environment that OpenAI explicitly cited in its pause — was always a structural challenge the September 2025 announcement did not address.

The broader context here is one of political sequencing. The UK government approached Nscale and OpenAI shortly before Trump's visit, asking them to commit to Cobalt Park in time for the announcement. The commercial groundwork — site visits, engagement with local authorities, detailed energy procurement — appears to have been expected to follow. In this case, it largely did not.

That sequencing problem is not unique to Stargate UK. Governments routinely trade in headline investment figures that blend committed capital with conditional pledges and letters of intent. The issue is disclosure: when £20 billion of a £30 billion figure is characterised as hypothetical by the reporting, and confirmed FoI evidence shows key partners never visited the designated site, the gap between the announcement and the underlying deal is material.

OpenAI, Nscale, and the UK government have not publicly reconciled those gaps. Until they do, Stargate UK reads less as a paused project than as an announcement that outran its own foundations.