Setback for UK data centre shows flaws in AI strategy

OpenAI’s decision to “pause” the development of a massive data centre in the AI Growth Zone in North Tyneside’s Cobalt Park is a blow to Britain’s ambitions to be an artificial intelligence superpower.

The project was promised during Donald Trump’s state visit to the UK last September, but the company now says the “right conditions” are not in place for “Stargate”, pointing to high regulatory and energy costs.

Cutting red tape and accelerating planning approvals for data centres is key to the government’s AI Growth Zones strategy. Is the reality not yet matching the ambition? Or is OpenAI trying to strong-arm the government into weakening intellectual property protections on content that it wants to use to train its large language models?

The high cost of energy in the UK and the risks arising from its dependence on foreign fuels – highlighted by the current situation in the Middle East – are a potential existential weakness in the AI strategy. It can cost three times as much to run a 500-megawatt data centre in the UK as in the US or the Nordic countries.

The government’s preferred option, next-generation nuclear power, could change the economics dramatically, but it will not start to come on stream before the mid-2030s. The global scale-up of AI data centres is happening now.

The government needs to either dramatically shift its short-term energy strategy or pivot to a different AI strategy, focused not on building massive computing power but on becoming the world leader in small language models.

These use far less energy. Britain could build AI small language model clusters in sectors where the country already has industrial strength, such as finance, law, medical research and applied materials.

Photograph by Wilf Doyle/Alamy