A Guardian investigation has revealed that OpenAI apparently never visited a key site designated for its Stargate UK data centre project before the initiative was publicly announced. This revelation deepens concerns that the UK government's flagship AI infrastructure programme was built on press releases rather than thorough due diligence.
Stargate UK, a partnership between OpenAI, Nvidia, and British cloud provider Nscale, was unveiled in September 2025 during Donald Trump's state visit to the UK. The project promised up to 8,000 Nvidia GPUs deployed at sites in north-east England by the first quarter of 2026, with potential scalability to 31,000 GPUs over time. It was part of a broader £31 billion package of technology investments that the UK government presented as proof of Britain's AI competitiveness on the global stage.
The Scaffolding Yard: Reality vs. Promises
The Guardian had previously investigated the physical reality of the project in March 2026, when reporter Aisha Down visited the site of Nscale's planned supercomputer in Loughton, Essex. What she found was a functioning scaffolding yard with no evidence that construction had begun on a facility that Nscale had claimed would be operational by the end of 2026. Land records at the time showed no evidence of Nscale's ownership of the site, with the property still registered to a different company entirely.
Nscale had announced in September 2025 that it had “confirmed” its UK investment by purchasing the Loughton site and promised 23,040 Nvidia GPUs running by early 2027. Yet, months later, the site remained unchanged, raising serious questions about the due diligence behind such high-profile announcements.
The Unsigned Contract: Government Without Auditing
The March investigation also uncovered that the UK government had issued a press release describing a £1.9 billion investment contract with Nscale. However, no such contract had ever been signed. When questioned, the Department for Science, Innovation and Technology admitted it was “not playing an active role in auditing these commitments.” This lack of oversight is particularly troubling given the government's reliance on these figures to demonstrate economic growth and technological leadership.
The government's AI Opportunities Action Plan, published in January 2025, cited £14 billion in private commitments to UK AI infrastructure. These figures were based on company self-reporting rather than audited disclosures, a method the Guardian described as producing “phantom investments.” This pattern of unverified pledges raises the risk that the UK's touted AI boom is more rhetorical than real.
The Pause: Energy Costs and Regulatory Hurdles
OpenAI formally paused Stargate UK in April 2026, citing the high cost of industrial electricity in Britain, which is roughly four times higher than in the United States, Finland, Norway, and Sweden. The company also pointed to unresolved regulatory questions surrounding AI and copyright. A spokesperson stated that the company would “continue to explore Stargate UK” when “regulation and the cost of energy enable long-term infrastructure investment.”
Notably, no planning applications had been lodged and no construction had begun at Cobalt Park, the primary Stargate UK site near Newcastle. This stands in stark contrast to the project's grand announcement, reinforcing the gap between promises and execution.
Where the Money Went Instead: Portugal and Cheaper Energy
While Stargate UK stalled, Nscale invested €695 million in Portugal to supply 66,000 Nvidia Rubin GPUs to a Microsoft data centre campus in Sines. The company, which reached a $14.6 billion valuation in two years, found more favourable energy prices and planning conditions on the continent. This shift underscores the competitive disadvantage the UK faces in attracting large-scale AI infrastructure investments due to high industrial electricity costs and regulatory complexity.
Nscale's smaller UK project with BT and Nvidia, providing up to 14 megawatts of sovereign AI capacity across three existing BT sites, is proceeding. However, its modest scale is a fraction of what Stargate UK had promised, highlighting the gap between the government's AI ambitions and what the market is willing to actually build in Britain at current electricity prices.
The Pattern: Counting Investments That May Not Materialise
The UK government continued to announce large AI investment figures at London Tech Week in June 2026, including pledges from AMD and Nebius. Whether those commitments will face the same level of scrutiny as Stargate UK remains to be seen. The Stargate UK episode exposes a structural weakness in how the UK counts AI investment: the government tallies pledges at the announcement stage but does not verify them, creating a gap between press releases and actual data centres that can be measured in years—or as a separate report on the UK's AI-climate conflict warned, may never close at all.
The broader context of the UK's energy market is critical here. The country's industrial electricity prices are among the highest in the developed world, largely due to network costs, carbon taxes, and renewable energy subsidies. For data centre operators, who consume enormous amounts of power, this cost is a decisive factor. Finland and Norway, by contrast, offer abundant hydropower at low prices, making them attractive alternatives. Sweden benefits from a well-developed grid and low-cost nuclear and hydro capacity.
Regulatory uncertainty around AI and copyright also plays a role. The UK government has yet to pass comprehensive legislation addressing AI training data, liability, or governance, deterring firms that require long-term stability. In contrast, the European Union has implemented the AI Act, providing a clear framework that, while strict, offers predictability. The UK's piecemeal approach may be undermining its ambitions.
The Stargate UK episode also echoes historical patterns in UK industrial policy, where grand announcements often fail to materialise. From the cancelled Swansea Tidal Lagoon to delayed nuclear projects, the UK has a track record of promising large infrastructure but struggling to deliver due to a combination of high costs, regulatory hurdles, and lack of government enforcement. AI data centres are simply the latest sector to face these challenges.
Looking ahead, the UK government may need to reconsider its approach. Rather than relying on self-reported figures from companies that may not follow through, it could implement a system of verifiable milestones, conditional tax incentives, and direct oversight. The cost of energy also requires a long-term strategy, such as investing in grid capacity, expanding renewable generation, and considering targeted industrial subsidies for high-value sectors like AI infrastructure.
For now, the Stargate UK project stands as a cautionary tale about the gap between 'AI nationalism' rhetoric and the practical realities of building hyperscale data centres. The scaffolding yard in Loughton remains a working yard. The government's headlines remain unfulfilled. And the UK's AI ambitions, at least for large-scale computing, remain on hold.