The Guardian has examined government plans to build Britain’s AI infrastructure for the future, finding some of these to be, in the words of one source, at best unclear and at worst “complete bunk”.
The plans in question are for AI growth zones, which are supposed to be regions where the government supports companies to build massive AI datacentre complexes, of 500MW or greater. These could be bigger than any now operating in the UK.
Ambitious plans were announced, but some of them are plainly not feasible and the government appears to have simply ignored concerns about how Britain is going to execute some of its most ambitious infrastructure projects in years.
Here is what the investigation found:
What is an AI growth zone and what is required to build one?
AI growth zones are areas where the government wants to encourage the building of AI datacentres. The policy, as executed, appears to be an attempt to mash together two goals: creating jobs in former industrial areas and ensuring Britain stays ahead in the AI race.
The government has announced five AI growth zones so far. One of these is in Lanarkshire, east of Glasgow. Another is in North Tyneside, near Newcastle – this was the site of the Stargate UK project.
What was supposed to happen was that local authorities and companies were to apply to the government to become AI growth zones. In these applications, they were to demonstrate that they could build 500MW AI datacentres by 2030. This meant submitting evidence that their plans were technically feasible.
Bonus points if the sites could also demonstrate that they had investment and could create jobs.
Are the plans to build AI growth zones feasible?
At least in the case of Lanarkshire, the most recently announced AI growth zone, no.
The facility – as announced – is supposed to be powered by a massive amount of on-site renewable energy, and not Britain’s electricity grid. The site’s developer, DataVita, has described plans to power it that would amount to building the UK’s largest onshore windfarm in the next four years.
But that is not the real plan: in internal communications, and then in response to the Guardian, DataVita and the government acknowledged that the site would just connect to the grid.
Meanwhile, the Guardian examined if DataVita could actually build that amount of renewables in the first place, given the land it has in planning applications. The answer is very close to no, as DataVita appears to currently have roughly a 10th of the land it would need.
What happened with Stargate UK?
The Guardian also examined some plans behind Stargate UK, the AI growth zone in North Tyneside from which OpenAI recently withdrew.
It found that plans for the site appeared to come together in the short period before the US president Donald Trump’s visit to the UK last year, driven by political goals. As one source put it, the government needed “a big announcement”.
This meant that OpenAI and Nscale, which were supposed to build the site, were not involved in the application process. This in turn raises questions over how thought out the plans to build it actually were. Local authorities have suggested that Stargate UK did not have the necessary grid capacity and infrastructure for a project of its scale.
What about the money and jobs?
The job numbers for Lanarkshire appear inflated; they are taken from industry estimates for another British datacentre, then increased because Lanarkshire is a bigger site. The government has said the datacentre will create 3,400 jobs; a Scottish charity has suggested the real figure will be hundreds of times smaller.
The money that the government has said will go into a community fund for Lanarkshire, £543m, is not in place. Rather, it is to come out of the revenues that DataVita generates – if it generates them.
Meanwhile, the figure that the government gave for potential future investment in Stargate UK, £20bn, appears to be circular, in that the government has said the site will have that amount because it requires that amount for it to be built.
What does this mean for the UK’s AI ambitions?
Electricity is not a triviality but one of the deciding factors in whether AI infrastructure can be built. It is a big question everywhere – datacentres are standing empty in California because they cannot get a connection. Europe has recently taken a stab at addressing the power issue in its own AI law.
Power is an even bigger question in Britain, where the grid is under increasing strain, and there is an eight- to 10-year wait for a connection. Giving a massive datacentre an expedited connection means, potentially, letting it cut the queue ahead of houses and hospitals. AI growth zones were not supposed to involve this kind of decision-making.
The fact that the government appears to have in these cases overlooked the central issue that will determine whether its AI developments can be built is not a positive sign.
