NextFin News - James Wise, chair of the UK's £500 million ($675 million) Sovereign AI Fund, has urged ministers to "think like an investor" and use government procurement as the primary tool for growing domestic AI startups, arguing that Britain has failed to deliver the kind of anchor contracts that turned US firms such as SpaceX and Palantir into "very valuable companies." The intervention, in an interview published on Wednesday, places procurement reform - not fresh capital - at the centre of the government's AI growth strategy, coming days after Chancellor John Healey unveiled a £100 million procurement scheme for homegrown AI companies at the G20 finance ministers' meeting in North Carolina.
The tension Wise identifies is structural rather than financial. The UK is home to 5,862 AI companies employing 86,139 people, a sector that generated £23.9 billion in revenue and £11.8 billion in gross value added in 2024, according to the government's own AI sector study. British startups raised $7.8 billion in venture capital in the first quarter of 2026 alone - the strongest start to a year since 2022. Yet the share of scale-up funding coming from domestic investors sits at around 15%, one of the lowest rates in Europe. The government has built a £500 million fund that looks like a venture capitalist on paper. The question now is whether Whitehall can behave like one in practice.
The fund that was built to act like a VC - and its bigger ask
The Sovereign AI Unit was launched on April 16 by Technology Secretary Liz Kendall as a "first-of-its-kind" national effort backed by the Department for Science, Innovation and Technology. From the outset it was designed to differ from traditional grant-giving bodies: it operates as a state-backed venture capital fund with a mandate to generate commercial returns, offering equity investments of between £1 million and £10 million per company, up to 1 million GPU hours on the UK's AI Research Resource supercomputer network, visa decisions within one working day plus an initial ten cost-free visas for top R&D talent, and hands-on help navigating regulation, data access and government procurement.
By government standards the fund has moved quickly. Its first equity investment went to Callosum, a London-based AI infrastructure startup founded by Cambridge PhDs Danyal Akarca and Jascha Achterberg that builds systems allowing different types of computer chips to work together efficiently. Six further startups received access to supercomputing capacity at launch, and as of July 2026 the fund had backed four companies through direct equity. Alongside the equity vehicle, Sovereign AI opened a £282 million Strategic Assets Grants Programme to fund shared inputs for the wider ecosystem, such as high-value datasets and automated laboratory infrastructure.
Yet Wise's argument is that capital is not the binding constraint. The failure he describes is that Britain has not acted as the first customer for its own startups - the role that US defence and space procurement played for SpaceX and Palantir. In his view, the US model worked because the state became a demanding, referenceable buyer long before private markets were willing to write large cheques.
Innovation in US government procurement helped make them "very valuable companies."
James Wise, chair of the UK's Sovereign AI Fund and a partner at Balderton Capital, was pointing to a mechanism that British policy has so far missed. Both SpaceX and Palantir built their early revenue bases on public-sector contracts before expanding commercially, a path European founders have far less often been able to follow.
The timing of the remarks is deliberate. On August 31, Healey announced the first procurement competitions under a new £100 million Sovereign AI R&D Procurement Scheme, aimed at domestic AI startups working on public services - from cutting NHS waiting lists and improving patient care to strengthening cyber and national security. The Chancellor framed it as part of a broader promise: the scheme is "the latest part of Sovereign AI's support to help promising British AI start-ups start in the UK, scale in the UK, and win globally." He also confirmed that the UK plans to open its AI Economics Institute to international cooperation with G7 countries, extending the policy push beyond funding into the evidence base on AI's economic effects.
The question the fund now faces is whether procurement - an area governed by accountability rules, value-for-money tests and multi-year budgeting - can move at the pace of a startup. It is the difference between a state that writes cheques and a state that behaves as a demanding, referenceable customer.
Why demand, not capital, is the binding constraint
The UK's funding gap is real, but it is concentrated at the scale-up stage rather than at seed. In the first quarter of 2026, British startups raised $7.8 billion in venture capital, up 60% from the same period a year earlier, with megadeals accounting for $5.1 billion of the total. The problem is who writes those later-stage cheques: the share of investment at the $100 million-plus scale-up stage coming from domestic investors is around 15%, one of the lowest rates in Europe. British founders routinely reach Series B and beyond only by selling to overseas funds, which shifts decision-making, intellectual-property ownership and eventual exit value abroad.
Average fund size tells a similar story. In 2025, UK venture funds raised an average of £83.2 million, compared with $122.8 million - approximately £92.6 million - for their US counterparts, a gap of only about 11%. The capital base is not dramatically smaller; the domestic willingness to write large, conviction cheques is. Venture capital accounted for 46.1% of UK startup funding by value in 2025, with corporate and corporate venture capital investors contributing 24.6% - meaning the majority of capital still comes from financial investors rather than strategic customers who can also provide revenue.
This is where the procurement argument bites, because a government contract does three things a grant cannot. First, it provides revenue rather than subsidy, validating the startup through a paying customer with a real operational problem. Second, it is referenceable - a ministry deployment becomes a case study that de-risks the company for the next private investor. Third, it imposes disciplines that grant-funded research rarely does: the product must work in production, meet security standards and survive scrutiny.
The mechanism is a form of non-dilutive venture fuel. At the stage when equity is most expensive, a public-sector contract extends the runway, funds hiring and signals to private markets that the technology works outside the lab. The government does not need to pick winners in the way a grant committee does; it needs only to be a demanding first customer, and the market does the rest. That is the investor mindset Wise is describing: the state's job is not to fund everything, but to de-risk the part of the journey where private capital hesitates.
This reading is structural, not cyclical. A cyclical shortage of risk appetite reverts when interest rates fall or sentiment improves. What Wise is describing does not self-correct: it is the product of fragmented demand across departments, civil servants who are penalised for failed purchases but not rewarded for innovation, and procurement rules designed to prevent waste rather than to create companies. The evidence sits in the numbers - the 15% domestic share of scale-up funding has persisted since 2025, and the fund-size gap has held through multiple rate cycles. A cycle reverts on its own; a structure does not. That is why the argument is for a change in how the state buys, not merely a new budget line.
The multiplier: when £500 million is a lever, not a ceiling
The second-order implication of Wise's argument is that the headline fund size understates what is at stake. The Sovereign AI equity fund is £500 million. If the £100 million procurement scheme helps even a handful of startups reach the point where they raise private follow-on rounds, the state's £500 million is no longer the ceiling on British AI investment - it becomes the lever that unlocks several times that amount from private markets.
The UK already has the substrate for this to work. The technology sector's combined market value reached $1.6 trillion, according to the Tech Nation Report 2026, with AI companies accounting for 32% of that total - a share that has more than doubled over five years. The country is home to more than 2,500 venture-backed AI startups, and in the first half of 2026 a staggering 77% of the $14.5 billion in UK venture capital went to AI companies. UK technology startups overall raised more in that period than all other major European markets combined.
But concentration is the flip side of strength. AI's dominance of UK venture flows means the marginal startup working on public services rather than frontier models faces a thinner market. Public-sector procurement can fill precisely that gap, creating a demand signal in areas where commercial buyers move slowly. The NHS, local authorities and security agencies have problems that AI can address; turning those needs into contracted demand gives founders a market that exists regardless of venture sentiment.
There is also a geographic and ownership dimension. When a startup's first major customer is overseas, its engineering centre of gravity tends to follow. When the first customer is a British public body, the company has a reason to keep decision-making, tax base and intellectual property at home - which is the explicit goal behind the "start in the UK, scale in the UK, win globally" framing. The £100 million scheme is therefore not just industrial policy; it is a retention mechanism for a sector where relocation risk is constant.
The US playbook - and its British irony
The American model works because procurement and venture capital reinforce each other. Defence and intelligence contracts gave SpaceX and Palantir non-dilutive revenue when equity was scarcest, and the reference value of those contracts compounded with each subsequent private round. Britain has begun to test the same logic, though with an uncomfortable twist: in September 2025 the UK signed a strategic partnership with Palantir that could see up to £1.5 billion ($1.8 billion) flow into defence-sector innovation, and the company has made its software available through Crown Commercial Service frameworks. The model for what US procurement can build is itself a US company winning UK contracts.
The August 31 procurement scheme is the first serious attempt to redirect that logic toward homegrown firms. It is explicitly aimed at startups working on public services rather than at established defence primes, which matters: it creates a pipeline of referenceable deployments that a Series B investor can underwrite. Public-sector AI spending is already accelerating - the public sector has spent £1.41 billion on AI through early August 2026, surpassing the £1.18 billion awarded across the whole of 2025 - but the question is how much of that spend reaches domestic startups rather than incumbent suppliers.
The appointment of experienced venture investors to Sovereign AI's leadership is a direct attempt to close the capability gap. Wise is a partner at Balderton Capital; in May 2026 Suzanne Ashman, a former General Partner at LocalGlobe and Latitude, was appointed managing partner of the fund's investment committee, with Joséphine Kant serving as head of ventures. A handful of seconded investors cannot rewrite the procurement habits of the entire civil service, but they can set the standard for how the fund itself behaves - and, if the procurement scheme is run with similar discipline, how departments learn to buy.
The counter-thesis: procurement is slow by design
The strongest case against Wise's argument is that it underestimates why procurement is slow. Public bodies are accountable for every pound, and the officials who sign contracts face personal scrutiny if a deal goes wrong but receive no reward if it goes right. Startups live or die on speed. A company that can pivot in a week cannot wait eighteen months for a framework agreement, and no amount of "investor mindset" exhortation changes the accountability structure underneath it.
Venture capitalists also have a tool governments lack: they can walk away. A state-backed fund with a political mandate to "back Britain" faces pressure to keep supporting national champions even when the thesis has broken. Sovereign AI's commitment to operating "like a top-tier VC firm" and generating commercial returns is the guardrail against this, but it will be tested the first time a portfolio company fails and the headlines ask why taxpayers backed it.
There is also a capacity question. Writing equity cheques of £1 million to £10 million and deploying up to 1 million GPU hours per startup requires investment professionals who can evaluate frontier AI - a scarce skill set inside government. The senior appointments address this at the top, but the procurement scheme will live or die at the working level, where category managers and commercial directors decide what "value for money" means.
The falsifying signal is concrete. If, twelve months after the August 2026 launch, fewer than ten startups have won contracts under the £100 million procurement scheme and none has raised a follow-on private round citing a government contract as validation, the "anchor customer" thesis has not taken hold. Equally, if more than half of the £100 million remains unspent by the end of 2027, the investor mindset has not penetrated Whitehall's buying behaviour. Either outcome would suggest the constraint is not mindset but machinery.
What to watch: three horizons
In the short term, the signal is deployment speed. The first procurement competitions were launched on August 31; how long they take to award, and whether the winners are genuine startups rather than incumbent suppliers rebranded, will set the tone for the whole initiative. Fast awards suggest the new mindset is real; slow ones suggest the old machinery has absorbed the reform.
Over the medium term, the key metric is follow-on capital. The Sovereign AI Fund's success should not be measured by how many companies it backs, but by how many of them raise private money afterward - and from domestic investors. A sustained rise in the 15% domestic share of scale-up funding would be the clearest evidence that the anchor-customer mechanism is working. Watch also whether the £1.41 billion of public-sector AI spend begins to flow toward smaller domestic suppliers rather than concentrating with a handful of large contractors.
In the long term, the question is structural: does Britain produce a generation of AI companies that scale at home rather than relocating? The base case is that procurement reform proceeds slowly but measurably, with the £100 million scheme partially deployed and a handful of visible successes by 2027. The upside case is that a few high-profile deployments - in the NHS or cyber security - create a self-reinforcing pipeline of referenceable contracts and private co-investment, lifting domestic scale-up participation toward European averages. The downside case is that accountability rules and departmental fragmentation slow awards until the startups have already raised overseas or folded, leaving the £500 million fund writing cheques into a market whose demand side never materialised.
Wise is right that procurement is the deeper lever, and the £100 million scheme is the right instrument. But a £500 million fund and a £100 million procurement pot are not, on their own, an industrial strategy - they are a down payment on a change in how the state buys. Britain's AI ambitions will be decided less by how much money the government writes than by whether it can become the kind of customer that makes startups investable.
Explore more exclusive insights at nextfin.ai.
