OpenAI’s apparent failure to visit key site raises questions over UK investment

The story of “Stargate UK” was sold as a landmark in modern industrial strategy: a transatlantic partnership, a flagship datacentre, and a signal that Britain would not merely consume the next wave of artificial intelligence but help to host and power it. Ministers spoke in the language of momentum and scale. The north-east of England was presented as the setting for an economic pivot, with Cobalt Park in North Tyneside recast as an “AI growth zone” and a headline figure of £30bn in investment offered as proof that the country’s pitch to global tech had begun to land.

Yet the more closely the project is examined, the more it resembles a government announcement searching for the substance that should normally precede it. The plans were paused in April, with OpenAI pointing to regulatory uncertainty and the high cost of energy. Those are credible impediments for any operator contemplating a facility that consumes power on the scale of a small city. What is harder to reconcile is the suggestion, now unearthed through freedom of information material and corroborated by sources, that OpenAI does not appear to have visited one of the scheme’s central sites and that the supposed additional £20bn of “potential” investment was, at best, an arithmetical projection rather than money attached to identifiable partners.

There is a familiar pattern here, and it reaches beyond a single company or a single ministerial press release. Artificial intelligence has become a theatre in which governments perform modernity. The props are recognisable: a supercomputer, a cluster, a growth zone, a strategic partnership. The political reward is immediate, particularly when staged alongside diplomatic spectacle. The cost arrives later, when local authorities ask what has been agreed, when grid operators look for formal applications, and when companies quietly weigh whether the economics work at all.

Stargate UK was unveiled amid the polished choreography that accompanied Donald Trump’s state visit to London last September. It was framed as a counterpart, albeit a smaller one, to the Stargate AI project in the United States, where OpenAI has spoken of investing $500bn to bolster American leadership. In Britain the cast was more modest and more plausible: OpenAI working with the UK firm Nscale, which has ambitions to build computing infrastructure in Essex, and with Nvidia, the dominant supplier of the specialised chips that make modern AI systems possible at scale.

The north-east site carried the symbolic weight. Cobalt Park was the place the public could picture, the map pin that made an abstract promise look tangible. Ministers designated it an AI growth zone during the visit, a phrase that suggested the state could will an ecosystem into being through planning preference and political attention. But behind the ceremony, the normal sequence of events appears not to have occurred. A freedom of information response returned to the Guardian indicates that neither OpenAI nor Nscale met the local authorities in North Tyneside. Only Nvidia is recorded as having visited the North East combined authority, and that took place in February 2026, months after the announcement.

This is not a pedantic detail about diary management. Major infrastructure projects, particularly those that touch land use, energy supply, transport access and workforce planning, depend on early contact with the institutions that control the relevant levers. Even the most centralised investor cannot conjure grid connections, highways improvements or planning permissions out of press coverage. The absence of meetings suggests either that the project was never sufficiently developed to require them, or that engagement was left so late that the announcement arrived before the practical groundwork.

Sources close to the early discussions describe a process driven by political need rather than industrial logic: a desire for a large, photogenic commitment to accompany a presidential visit. One source characterised it as a search for “a big announcement”. Another described Nscale as being taken unawares, effectively instructed to align itself with a government narrative. Such accounts should be treated cautiously, as all parties have reputational interests in where responsibility lands when a high-profile plan stalls. But the available documentary picture, and the gaps within it, point in the same direction: a centrepiece of “US-UK AI cooperation” that may have existed primarily as messaging.

The money is where the rhetoric most clearly outruns the evidence. The government’s own press release said the growth zone was “set to” bring in £30bn. Within that figure, £10bn was described as “committed” by Blackstone, which is developing a separate datacentre project in the region and which appears still to be progressing. The remaining £20bn was presented as “potential” investment from “future partners”. When pressed on how that number had been calculated and who the future partners might be, the government declined to give detail, saying only that it reflected the total investment the site could attract.

Spotlight on Corruption, which pursued the question, received a clearer explanation, and it was revealing. The figure was presented as the amount of money the site would need to build a datacentre and to obtain the computing power required to make use of its electricity supply, described as 1.1GW. In effect: the site might attract £20bn because £20bn is what a project of that notional size would cost. That is not the same as having investors prepared to spend it. It is an estimate dressed as a commitment, and it is the sort of linguistic ambiguity that can energise a news cycle while leaving communities with expectations that have no firm basis.

Kamila Kingstone, a senior campaigner at Spotlight on Corruption, called it disingenuous to imply the £20bn would be forthcoming when it reflected the amount needed. Her criticism is aimed at the gap between the language of certainty and the reality of risk. Communities are asked to believe that “eye-watering” sums are on their way, and local leaders are expected to align skills programmes, planning assumptions and political capital with the promise. When the promise evaporates, the reputational damage is not evenly shared. National politicians move to the next announcement; local authorities live with the residue.

John Johnsson, the Conservative leader in North Tyneside, described local surprise at the Stargate UK announcement. “We were really, really taken aback,” he said, noting that officials were not made aware of discussions and were confronted with “pizazz” rather than prior coordination. His remarks speak to a practical truth: a press release can flatter a place while simultaneously bypassing it. When a project is real, the locality is usually a partner early on, not a spectator learning of its own future from a national media moment.

There were also, even at the level of basic engineering, reasons to doubt the timetable implied by the launch. A separate freedom of information response from the National Energy System Operator suggested that the site did not have a grid connection. Instead, it submitted an alternative solution to power itself, though the details were redacted. The point is less the specifics of that proposal than the fact it existed at all. Data centres of the scale discussed are not simply buildings with servers; they are, fundamentally, power projects. If grid connection is uncertain, the business case can collapse quickly, particularly in a country where energy prices have been volatile and where new transmission infrastructure is politically contested.

OpenAI’s own explanation for pausing the plans was that regulation and energy costs did not yet support long-term investment. That statement, repeated when the company was asked about site visits, is carefully constructed. It keeps the door ajar, affirms “huge potential” and avoids an outright repudiation of Britain. It also provides a ready-made rationale that does not require the company to litigate the details of government expectations or internal decision-making. Nscale, for its part, said its chief commercial officer had gone to North Tyneside but did not clarify whether any meetings took place, and the public record does not show them.

The bigger question is not whether a single site meeting happened, but what the episode tells us about how Britain is trying to compete in the AI infrastructure race. Data centres, chip supply and energy are now the hard power of the digital economy. Countries able to host large clusters gain not only jobs but strategic leverage. They can attract research talent, court downstream industries and, crucially, promise the computing capacity that modern companies require. For governments, the temptation is to speak as if such capacity can be announced into existence, as if a designation of “growth zone” or a handshake between leaders is itself a form of infrastructure.

In reality, the constraints are stubborn. Planning systems are slow. Grid connections are finite. The lead times for specialist equipment are long. And the economics of AI are shifting: training frontier models consumes colossal power, while the commercial returns remain uncertain and concentrated among a small number of firms. If ministers want Britain to host that future, they must do more than court headlines. They must make the long, sometimes unpopular decisions that infrastructure demands: accelerate grid upgrades, clarify regulatory approaches, align energy pricing and land policy, and accept that not every promised “mega-project” will survive contact with cost.

The Stargate UK story also sits alongside a wider investigation into “phantom investments” in the government’s AI drive. That term captures an uncomfortable reality: political systems reward announcements, not delivery. A promise of billions can be more valuable than the incremental work of securing planning consent, negotiating power supply, and building local training pipelines. Over time, however, a credibility deficit accumulates. Investors become wary of being used as stage dressing. Local authorities become sceptical of central claims. And the public begins to treat industrial policy as branding rather than governance.

None of this means the north-east cannot host serious AI infrastructure. Britain has advantages: a respected research base, established legal institutions, and a cluster of technology firms in and around London, Cambridge, Oxford and increasingly other regions. The question is whether those strengths can be paired with the physical necessities of the AI era. Ministers now say a taskforce co-chaired by the technology secretary and the North East mayor, Kim McGuinness, is driving planning, investment and skills, and that the growth zone will increase its energy capacity to 1.1GW once fully operational, with more than 400MW expected to come online in 2028. Those are the sorts of numbers that, if matched with deliverable projects, could begin to restore confidence.

But confidence will not be rebuilt by repeating the language of potential. It will be rebuilt through specificity: who is investing, what is being built, what approvals have been secured, how the power will be delivered, and what the timetable actually is. The private sector will always retain the right to change its mind, particularly in a fast-moving industry. The state cannot guarantee that OpenAI, or any single company, will choose Britain. What it can do is create conditions where the next company that looks seriously at building here finds a country prepared for the practicalities, not merely the optics.

For now, Stargate UK reads as a cautionary tale about the difference between being photographed alongside the future and funding it. The pause may yet become a delay; the partners may yet reassemble; the economics may yet shift. Even so, the episode exposes a vulnerability at the heart of Britain’s AI ambitions: the risk that policy becomes a series of declarative acts, each seeking to borrow credibility from global brand names, while the underlying infrastructure remains unresolved. In an industry where power, land and time are as decisive as algorithms, that is not a communications problem. It is the central challenge.