Summary

• The UK AI Hardware Plan sets out support across innovation, skills, procurement, and investment.
• Measures include a heterogeneous AI supercomputer, specialised chip procurement, AI hardware innovation funding, and semiconductor skills support.
• The plan aims to connect UK chip design, photonics, hardware security, procurement, and AI infrastructure into a more coherent industrial pipeline.

The Department for Science, Innovation and Technology has published a UK AI Hardware Plan, setting out how government wants to strengthen the chips, advanced materials, skills, procurement routes, and investment pipelines behind the country’s AI economy.

The plan puts targeted public and private support behind AI hardware across four pillars: innovation, skills, procurement, and investment. It is intended to help UK companies develop, demonstrate, deploy, and scale technologies that support AI systems, including specialised chips, photonics, edge inference, hardware security, and energy-efficient compute.

The largest procurement measure is a £750 million heterogeneous AI supercomputer for the AI Research Resource. The system is intended to integrate different advanced compute technologies, including specialised architectures and, over time, quantum computing, into a platform used on real research workloads.

Within that programme, the government is creating a £400 million procurement opportunity for specialised chips. That includes an expanded £150 million advanced market commitment to buy novel high-performance inference chips, followed by a further £250 million procurement for additional specialised hardware.

Other interventions include £120 million for AI hardware innovation, an £18 million hardware security research and development programme, a refocused Semiconductor Catapult, £80 million for semiconductor and AI hardware skills, and a new deeptech hardware venture fund led by Playground Global, backed by up to £150 million from the British Business Bank. The plan also links AI hardware companies to the £500 million Sovereign AI Fund.

The government’s AI Hardware Plan reflects a practical constraint on AI policy. Model development and adoption depend on physical systems: chips, interconnects, memory, power electronics, sensors, advanced materials, data centres, and people able to build and integrate them.

The UK has strengths in chip design, semiconductor IP, photonics, advanced materials, and hardware security, but has often struggled to turn research advantages into scaled industrial companies. The plan is aimed at reducing the gaps between early-stage innovation, testing, procurement, investment, and commercial deployment.

Public procurement carries much of the weight. By using the AI Research Resource and National Cloud Infrastructure Programme as routes to market, the government is trying to create early demand for UK-developed hardware. That could help companies prove performance on real workloads and attract private capital, provided procurement is fast, technically credible, and open enough to avoid becoming a slow grant substitute.

The hardware security element is also notable. As AI systems move into sensitive business and public-sector environments, vulnerabilities at the chip and system level become part of the trust problem. The plan’s focus on secure hardware signals that adoption depends not only on model assurance, but on confidence in the underlying compute stack.

Scale will be the difficult part. The global AI chip market is dominated by a small number of enormous companies with supply-chain advantages, manufacturing access, and developer ecosystems. The UK is not attempting to replicate the whole semiconductor supply chain, but to build leverage in areas where it has distinctive capability. That is a more realistic objective, although it demands disciplined execution.

If the plan works, UK firms could capture more economic value from the growth of AI infrastructure rather than simply importing the hardware needed to run it. If it fails, Britain risks remaining strong in research and early design while the industrial value of AI hardware scales elsewhere.