Five years after relocating his AI startup Gravitee to London, Rory Blundell has been spending his time elsewhere: house-hunting in Denver, Colorado. The CEO, who founded Gravitee in France a decade ago, has grown his agentic AI company to 130 staff and raised $60m this year. But he says the pull from the US is becoming hard to ignore: “All of our investors, everybody wants us to relocate to the US,” he says.
Gravitee is exactly the sort of company that UK ministers say they want to build. It is technically ambitious, scaling quickly, and rooted in Britain’s top-tier AI research base. London helped turn Blundell into a successful founder, but global competition means he is now weighing up whether the next stage of the company’s life should happen somewhere else.
Blundell’s dilemma is not unique. New data from The Observer Global AI Index suggests a broader pattern: the UK trains world-class AI talent and attracts strong investment, but struggles to turn those strengths into long-term companies and careers. That’s because foreign countries, mainly the US, offer easier access to three key startup ingredients: capital, AI infrastructure and talent networks.
“Many AI startups start in the UK but end up scaling in the US because that’s where hiring, capital, and customers converge,” says Nathan Benaich, founder of AI-focused venture firm Air Street Capital.
‘We see a consistent pattern: PhDs, post-docs and early-career engineers train here and then leave for the US, often permanently’
Nathan Benaich, Air Street Capital
On paper, the UK should be an AI winner. The index ranks Britain third in the world for its computer science education, behind only the US and China. Its research base is deep and well-funded: spending on AI and science-specific research and development sits among the highest globally, and last month the government pledged £137m to an “AI for science” strategy alongside heavy investment in AI-focused compute capacity. The UK also hosts the third-largest concentration of top AI research scientists in the world, according to the index, just behind China and the US.
But talent alone cannot build an AI industry. “The UK is uniquely good at producing world-leading research and uniquely bad at spinning it out into companies,” says Benaich, who says part of the problem lies in the way Britain commercialises the science it produces.
Benaich points to aggressive equity stakes taken by universities, and drawn-out negotiation cycles that can drag on for a year or more, pushing founders to look elsewhere.
According to Tom Hurd of AI research firm Zeki Data, Stanford University is “miles ahead” of Oxford or Cambridge by the number of companies that emerge from AI labs. UK government-backed AI institutes have absorbed significant public funding but are seen by some investors as under-delivering relative to counterparts in Canada or Germany. The Alan Turing Institute, once the UK’s hope to lead the world in AI, has been in crisis, with its CEO due to step down after a staff revolt earlier this year.
A bigger problem, says Carolyn Dawson, chief executive of startup-backer Founders Forum Group, is AI companies struggling to access capital. UK AI firms attracted more than £2bn in private funding in 2024, the index shows – about twice France’s (£1.1bn) and five times Germany’s (£440m) – but that still amounts to just 3% of private investment in the US.
Growth capital for frontier AI – the large-scale models and infrastructure that now dominate the sector – is overwhelmingly American. Founders who want to build and scale quickly discover that the deepest pockets are on the other side of the Atlantic, and feel they have to move to where the capital lives.
Without that crucial capital, the UK is starting to lose its biggest strength – its talent. “We see a consistent pattern: PhDs, post-docs and early-career engineers train here and then leave for the US, often permanently,” says Benaich. The index finds that Britain now retains fewer than half of its best academic AI researchers, whereas in France and South Korea, which are at similar levels of AI development, between 60% and 70% of researchers stay in the country. The US, meanwhile, keeps about 80% of its top academic talent pool.
Salaries play a large part. Senior AI engineers in the US can earn up to $500,000 a year, compared with about £150,000 in the UK, venture capitalists say. “The US, China and parts of Europe are actively headhunting AI researchers and engineers, offering enormous salaries and signing bonuses,” says Dawson, adding that one in three founders surveyed see access to top talent as their single biggest barrier to growth.
The final piece of the puzzle is infrastructure. Britain has some of the highest electricity prices in Europe, and hosts only two of the 20 most powerful European supercomputers. That leaves startups and researchers struggling to access the kind of compute needed for cutting-edge AI work. “Frontier-scale AI training and evaluation clusters accessible to startups, researchers and private labs would close a big part of the gap with the US,” says Benaich. “France is already moving aggressively here.”
Founders and investors who spoke to The Observer say Britain can still turn its research strengths into lasting commercial success – and compete with the biggest AI players – but only if it responds much more aggressively. “The UK needs to address the capital gains taxation system to get founders to come here,” says Blundell, who argues that more generous treatment of startup equity and exits would make a meaningful difference to where entrepreneurs choose to base themselves.
Last month, the government announced a £100m AI “advance market commitment”, designed to provide “first customer” capital to British startups that need early, committed buyers for their products. Founders and investors see it as a welcome signal, but say it is only a small step towards matching the scale of US public and private support.
Beyond tax and early demand, they highlight a familiar set of priorities: clearer and more predictable AI regulation; fast-track visas for top technical talent; procurement rules that make it easier for startups to sell into the public sector; and access to affordable, large-scale computing power. “The UK should be the Switzerland of compute and AI industrial strategy: neutral, fast, and predictable,” says Benaich.
Those changes, they argue, would play to advantages London already has. Its cultural pull, time zone and dense network of researchers and founders keep many in the city longer than the spreadsheets alone would suggest, and give Britain a chance to build a different kind of AI hub – not just a clone of Silicon Valley.
The question is whether the headquarters, the intellectual property and the long-term value of firms like Gravitee will stay in Britain as they grow. Where founders like Blundell choose to build – in London, Denver or somewhere else – may be the sharpest test of whether the UK can turn its AI strengths into companies that stay.
Additional reporting by Joe White and Hannah Schuller
Illustration by Observer Design