Skip to content
hw.dev
hw.dev/signal/uk-ai-hardware-plan-kendall-2026
SignalResultSense

UK Bets on a National AI Hardware Plan, Gallium Supply Chain Exposed

UK Technology Secretary Liz Kendall announced a national AI hardware plan targeting 5% of the projected $1T AI chip market, while the speech itself exposed Britain's critical gallium supply-chain vulnerability.

#ai-hardware#semiconductor#supply-chain
Read Original

UK Technology Secretary Liz Kendall used a Royal United Services Institute speech on April 29 to announce that Britain will develop a national AI hardware plan covering chips and semiconductors. The stated goal is 5% of a forecast $1 trillion AI chip market -- roughly $50 billion in revenue. The concrete deliverables: a yet-to-be-written hardware plan, a planned chips strategy, a GBP 500 million Sovereign AI Fund, and GBP 100 million already allocated for British chip startups.

The framing is reasonable as geopolitical argument. Five US firms now control 70% of global AI compute, up from 60% a year ago. The UK's only credible differentiation path is building evaluation and governance credibility through AISI -- the AI Safety Institute -- and then leveraging that into trade relationships with France, Germany, and Canada. Kendall explicitly pitched a 'middle powers' bloc as the diplomatic complement to the hardware plan. For hardware engineers, the interesting read is less the policy and more the supply chain exposure buried in the speech: China controls 98% of primary gallium and 83% of germanium, both essential inputs for compound semiconductors including GaN power devices and III-V RF chips.

The gap between rhetoric and physical reality is large. The UK's planned AI supercomputer is reportedly still a scaffolding yard in Essex. OpenAI paused a multi-billion-pound UK datacentre investment citing energy costs and regulation. AISI's evaluation work on Claude Mythos is genuinely impressive and Kendall correctly cited it as evidence that the UK can punch above its compute weight -- but model evaluation is not the same as semiconductor manufacturing.

For hardware engineers, the gallium dependency is the story that matters most. Every GaN power stage, every GaAs RF front-end, every InP photonics chip running AI datacentre interconnects has a supply chain that runs through Chinese gallium refiners. No national AI hardware plan that does not directly address that dependency is credible -- and Kendall's speech, admirably, named it even if it did not solve it.