TL;DR

The UK Sovereign AI fund has now invested in, or finished papering deals with, a dozen British AI companies — five made public and seven still closing. Between them those companies have raised £4bn. The disclosure comes in the third quarterly update from the fund’s chair, James Wise, published on 12 August.

Where the money has gone

Three names have surfaced since the last update. CuspAI works on models for designing advanced materials; Isomorphic Labs, the drug discovery business, was already known to the programme; and OLIX is building compute that uses light rather than electronics, aimed squarely at the memory and power ceilings that constrain GPU-based systems. Stage exposure runs from pre-seed through Series C, with more early-stage cheques expected as the portfolio fills out.

Thematically the fund is not spreading itself thin. It has concentrated on infrastructure and compute, on model architecture, on health and life sciences, on scientific discovery, and on trust and security — which is close to a list of the areas where a national position is defensible and a British company can hold something that is genuinely hard to replicate.

The interesting part is procurement

Beyond capital, the fund is widening access to the Isambard-AI machine, and to credits for commercial cloud, via the AI Research Resource. It has also been running founder sessions outside London — Glasgow, Newcastle, Manchester and Cardiff so far, Belfast due in October. The consequential change is a set of procurement challenges due shortly, in which British AI companies compete against problems named by government departments, with the fund itself acting as the paying customer.

That structure matters more than it sounds. The persistent complaint from UK AI founders is not capital scarcity but the absence of a first serious buyer, and public sector procurement has historically been where small British suppliers go to be slowly worn down. A programme that both invests and buys removes one of those obstacles.

Looking forward

Wise pairs the progress with a blunt risk list: export controls on frontier models, cyber capability rising on the other side of the ledger, British reliance on open-weight models originating in China, and American and Chinese spending on chips and data centres at a scale the UK will not match. The argument that follows is that domestic capability has become a security question rather than an industrial one.