ARIA (Advanced Research and Invention Agency) was set up in 2023 to fund high-risk, speculative research outside the cautious grant structures of UKRI. Modelled on US DARPA, it has an £800 million four-year budget and gives Programme Managers significant autonomy to close failing programmes.Source: https://www.aria.org.uk

What is ARIA's role in the UK government's AI compute strategy?

ARIA runs two compute-focused programmes: a £100 million Scaling Compute programme and a £50 million Scaling Inference Lab. Liz Kendall cited both in her April 2026 RUSI speech as instruments of sovereign control over AI infrastructure alongside the Sovereign AI Unit.Source: https://www.gov.uk/government/speeches/rebuilding-britain-for-the-new-world-liz-kendalls-speech-at-the-royal-united-services-institute

How was ARIA created and who funds it?

ARIA was created by the Science and Technology Act 2023 and is funded by the UK Government with an £800 million budget over four years. It is operationally independent, meaning ministers cannot direct individual funding decisions.Source: https://www.aria.org.uk

Background

The Advanced Research and Invention Agency (ARIA) is a UK Government funding agency established in 2023, modelled on the United States' DARPA. It was created by the Science and Technology Act 2023 and is operationally independent of government, with a mandate to fund research too speculative or long-horizon for conventional grants from UKRI (UK Research and Innovation). ARIA has a budget of £800 million over four years and is intended to make concentrated bets on transformative Science rather than spreading funds thinly. The agency's first Programme Managers, appointed in 2023–24, have backed computing, biology, and energy research.

ARIA has two computing-focused programmes directly relevant to UK AI: the £100m Scaling Compute programme and the £50m Scaling Inference Lab, both pre-announced before April 2026 and referenced in Secretary of State Liz Kendall's RUSI speech on 28 April 2026 as instruments of sovereign control over AI infrastructure. The Scaling Compute programme is designed to fund research compute at a scale unavailable through commercial providers; the Inference Lab focuses on deployment-side compute efficiency. Both sit alongside the Sovereign AI Unit (SAIU) in Kendall's framing as evidence that the UK Government's AI investment is a coherent portfolio rather than a series of disconnected grants.

ARIA's governance model gives it deliberate distance from Whitehall: Programme Managers operate with significant autonomy and are expected to close programmes that are not working rather than sustain them for political reasons. This creates tension with the AI Hardware Plan pre-announced for London Tech Week in June 2026, which requires ARIA's compute programmes to interlock with a ministerially-led hardware supply-chain strategy. Whether ARIA's autonomy survives that interlock will determine whether UK AI compute strategy is coherent or fragmented.

Source Material

Rebuilding Britain for the new world: Liz Kendall at RUSI Advanced Research and Invention Agency — official site

How the World Sees Them

UK Government

Kendall framed ARIA's compute programmes as sovereign-control instruments alongside the SAIU, pulling the independent agency into a ministerially-led narrative it was designed to be insulated from.

US comparison

DARPA's influence on US defence and AI industries (GPS, the internet, BERT) took decades; ARIA has four years of guaranteed funding and faces pressure to show results on a shorter political clock.

UK research community

ARIA's DARPA-inspired model offers researchers the high-risk, high-reward funding that UKRI's consensus-driven panels historically withheld; the AI Hardware interlock creates new accountability pressures.