The announcement, made on June 8, forms part of a broader £1.1bn AI Hardware Plan that the government says is designed to accelerate the UK’s deployment and scaling of AI technologies.

Of the total envelope, £750m will fund a new national AI supercomputer, with £400m of that allocated specifically to chip procurement, split between £150m for next-generation inference chips and £250m for more specialised processors as technologies mature.

The £400m commitment will see the government offer to purchase AI chips directly from UK-based companies, in an explicit attempt to encourage them to remain in the country, a priority Starmer framed with the slogan “start here, scale here and stay here.”

The plan sits within the wider £500m Sovereign AI programme announced by the Department for Science, Innovation and Technology in March, with the new supercomputer expected to be operational by 2030 and designed to integrate advanced processors and specialised AI accelerators alongside existing facilities such as Isambard-AI and Dawn.

The announcement landed alongside significant private sector commitments. AMD pledged up to £2bn and cloud infrastructure provider Nebius committed £1.7bn to build out AI capacity in the UK, underscoring the degree to which London Tech Week has become a staging ground for compute investment declarations.

On the same day, Scottish infrastructure firm Argyll Data Development announced the general availability of what it describes as the UK’s first sovereign AI inference cloud, built in partnership with Silicon Valley chip designer SambaNova.

The platform marks the first time SambaNova’s hardware has been made available through a general availability cloud service hosted entirely within the UK, giving organisations access to high-performance inference without routing data through infrastructure outside British regulatory jurisdiction.

The launch speaks directly to one of the thorniest constraints facing the sovereign compute agenda: energy. As governments and regulators across Europe introduce new measures to curb data centre power consumption and reduce pressure on national grids, Argyll’s platform takes a deliberate position on the infrastructure question.

SambaNova’s Reconfigurable Dataflow Unit architecture operates within existing UK data centres without requiring liquid cooling or specialist power infrastructure, an approach that, if it scales, could materially lower the energy overhead associated with bringing AI inference onshore.

Peter Griffiths, Chairman at Argyll Data Development, said the platform gives UK organisations “something they haven’t had before: premium AI inference running entirely within UK borders, at production scale, on existing data centres.”

Rodrigo Liang, CEO at SambaNova, framed the significance in infrastructure terms: “High-performance inference usually requires liquid cooling, specialist power, and purpose-built facilities, requirements that make sovereign deployment slow, expensive, and out of reach for most. Argyll has deployed that capability entirely within UK borders for the first time.”

RELATED STORIES

Nebius invests £1.7bn in UK AI push with Kao Data deployment

OpenAI pauses Stargate UK, citing energy costs and regulation