The request is straightforward, and the timing is not. British technology leaders are urging the government to commit more than £20 billion to sovereign AI infrastructure, at the exact moment the department that owned the UK’s AI strategy has been taken apart.
Prime minister Andy Burnham appointed Kanishka Narayan as minister of state for artificial intelligence with cabinet-level status. In the same reshuffle he dissolved the Department for Science, Innovation and Technology, created in 2023, and folded its functions into a new Department for Business, Innovation, Science and Trade under business secretary Jonathan Reynolds. The Times reported the £20bn call under the headline “Britain must commit £20bn to AI sovereignty, tech leaders urge”, and Computing covered the same campaign.
The concern behind the money is dependency. American hyperscalers host most of the models, chips and compute that British businesses rent, and the signatories’ argument, as the reporting puts it, is that reliance on US-owned AI infrastructure threatens national resilience. A country that cannot run its own models, in their framing, can be cut off or repriced at another government’s whim.
Industry reaction to the restructure split along predictable lines. Matt Harris of Hewlett Packard Enterprises welcomed the cabinet seat for AI while warning against losing momentum. Dr Andy Sellars of Cornerstone pointed to Narayan’s earlier work on chip strategy, including a £10 million semiconductor fund, as reason for optimism. Tom Glason of ScaleWise was blunter, telling industry coverage that founders and investors need stability rather than constant restructuring.
The practical mess is in the handover. The AI Security Institute has moved to the Cabinet Office. Governance of the £500 million Sovereign AI Fund, launched in April 2026 to back homegrown AI companies with capital, GPU hours and fast-track visas, is now split between the new business department and No. 10. Programmes that were designed to sit under one roof now sit under two.
What would £20 billion actually buy? On the evidence of the fund already running, the answer is compute first. The UK’s largest AI supercomputer project has slipped, with reports of power supply problems delaying installation, and the same constraint shapes every site plan in the pipeline. Data centre builds are now battles over grid connection queues as much as planning permission, which is why the AI minister’s public line, reported by Bloomberg, is that the UK must accelerate the build-out.
The budget on 28 October will not resolve a £20bn ask in one stroke. But a sovereign AI pledge that survives a departmental merger has to come with something the money can actually buy, and right now the clearest candidate is gigawatts.
