The UK's £100M scheme to buy AI from British startups, and why it matters

The UK’s Sovereign AI programme has moved from investing in British AI companies to buying from them. At the end of August the Sovereign AI Unit opened the first competitions in a £100M procurement scheme that invites British startups to build AI systems for public services, and the deadline for that first batch is 1 October. It is the most concrete version yet of an idea we flagged when the programme set out its plans over the summer: that the government wants to be its AI startups’ customer, not just their investor.

The shape of it matters as much as the money. The £100M is allocated to March 2030, with individual contracts running from £250,000 to £10M and most expected in the £1M–£3M range. The first four challenges are pointed at real problems: improving NHS productivity, making AI compute infrastructure more efficient, integrating AI across defence, and helping organisations manage AI risk. Crucially, the scheme is designed to strip out the things that normally keep startups out of government work: there are no minimum-turnover requirements, applications are streamlined, payments come up front rather than in arrears, each project gets a departmental champion from day one, and — the detail founders will care about most — companies keep ownership of their intellectual property. The unit is chaired by James Wise, with Josephine Kant as head of ventures; the Chancellor, John Healey, announced the launch at the G20 finance meeting.

Why this is a bigger deal than a typical funding line is worth spelling out. Public procurement is usually where good AI intentions go to die: multi-year timelines, turnover thresholds that exclude anyone young, and IP terms that make the work unattractive. For an early company, though, a signed government contract and a live deployment inside the NHS or defence is often worth more than a grant of the same size — it is revenue, a reference customer, and a credibility stamp all at once, the kind of thing that unlocks the next commercial deal. A state that removes the barriers it built and commits to spending on its own AI startups is doing something more useful than writing cheques.

It also lands in a changed landscape. This is the first procurement push since the machinery-of-government shake-up that abolished DSIT and moved AI into the Cabinet Office, and it fits the same through-line as the fund’s earlier equity bets: public money and public demand aimed squarely at the hard, strategically important layers of AI rather than the consumer apps that raise on their own.

The caveat, as ever, is execution. A scheme can be beautifully designed on paper and still stall if the NHS and defence buyers do not follow through, if the contracts are slow to sign, or if the sums are too small to compete with the pull of the US hyperscalers for the best companies. £100M across four years is real but not enormous, and a procurement scheme is only as good as the deployments it produces. But the design here is genuinely thoughtful, and the intent is unambiguous: the clearest signal so far that the UK means to be a buyer of British AI, not just a cheerleader for it. Whether it works is the thing to watch through the winter as the first contracts are awarded.