UK sets out £1.1bn AI hardware plan at London Tech Week

The UK's bet is that being a serious AI player in the 2030s means owning the metal, not just renting it from US clouds. At London Tech Week, the government unveiled a £1.1bn AI Hardware Plan on 8 June 2026, packaging sovereign compute, chip design and skills funding into one announcement, as the Guardian reported.

The headline number is £750m for a national supercomputer due by 2030, built on a heterogeneous architecture that mixes proven GPUs with experimental homegrown semiconductors. Sat alongside is £400m for chip procurement, with £150m earmarked to buy next-generation inference chips from British firms over summer 2026 and the rest going on more specialist parts as the technology matures. There is also £120m for a new AI Hardware Innovation Programme, and Playground Global is set to deploy up to £150m from the British Business Bank into UK-based AI hardware companies.

The political subtext is unsubtle. Britain has watched Graphcore go to SoftBank and Alphawave to Qualcomm, and the strategy openly targets 5% of a global AI chip market projected to reach roughly one trillion dollars in the early 2030s. Industry money turned up to validate the framing, with AMD committing up to £2bn over the next five years, and Nebius around £1.7bn to expand UK capacity using Nvidia systems.

The honest caveat is that £750m of supercomputer by 2030 does not put the UK in the same league as US and Chinese hyperscale builds, and 'homegrown semiconductors' is doing a lot of work in a country whose two best known AI chip designers have already exited to overseas buyers. What the reporting does not pin down is who specifically wins the £150m of summer inference orders, how the mixed-silicon architecture gets integrated in practice, or whether the £400m chip budget will be ringfenced to UK-designed parts. For UK chip startups and academic groups that have been surviving on grant scraps, though, the next twelve months suddenly look a lot more interesting.