Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
The UK government has insisted it will continue to fund pioneering artificial intelligence and supercomputing technology despite a backlash over Sir Keir Starmer’s shock decision to scrap support worth £1.3bn for projects it inherited.
The new Labour administration is scrambling to contain resentment over its decision earlier this month, which it made on the grounds that the investment had not been adequately funded by its Conservative predecessors.
Rescinded backing included £800mn promised to a major supercomputing project at Edinburgh university, an area where experts warn the UK is lagging behind its major peers.
The outcome will be an early test for Labour as it balances tough public spending constraints against its vow to boost science and technology to help hit ambitious economic growth targets.
“Supercompute is still a huge priority for the government — it’s a priority for [tech secretary] Peter Kyle and it’s definitely a desire of his that we get it right,” said one government official. “We want any government money to pack a punch. We are not abandoning supercomputing at all.”
Exascale supercomputing is the ability to produce a billion billion operations a second, called an “exaflop”. Considered the holy grail of computation, it is seen as crucial to widescale adoption of AI.
There are two such fully functioning exascale computers in the world, both in the US. Experts believe China also has at least one, though it has not submitted it to international leader boards on compute capacity. Large economies — from Japan to Europe and Saudi Arabia — are investing billions in similar projects.
Kyle told the Financial Times this week he was “gearing up for a bold approach” that would be based on an “AI action plan” currently being developed by entrepreneur and AI expert Matt Clifford due to be released in September. He said this plan “would set out our future compute needs and how we deliver them”.
Clifford said last year, before he took up a government role, that the “UK doesn’t have a very large amount of sovereign compute” and that it should “spend billions over the next decade to ensure the most important AI companies are being built in the UK”.
The UK has already invested £225mn in an AI-focused supercomputer being developed out of Bristol university, called Isambard-AI, which is set to be one of the most powerful exascale-class computers in Europe. Two other supercomputers — Archer2 at Edinburgh university and DiRAC, built across several universities — are due to reach the end of their lives in two years.
Two senior government insiders told the FT that Starmer’s ambition when it comes to the UK’s supercomputing potential had in no way been damped, but that there were simply some misgivings about the allocation of the pre-existing budget.
One senior government figure noted that the Edinburgh exascale project made little strategic sense given that it was not focused on AI, but on more traditional computing projects such as scientific simulations. The person added that they expected the new government to match the previous spending commitments in supercomputing in the forthcoming autumn Budget.
But shadow science minister Andrew Griffith told the FT it was a “misrepresentation” from Starmer’s government that the project was “unfunded”, given that it was always the case that investment would be accounted for in the 2024 autumn Budget. “It’s a huge setback for UK tech and . . . it seems a sign of the lowering of ambition when it comes to Britain’s plans to become a science and technology superpower,” he said.
Several supercomputing experts that the FT spoke to agreed that it was not the most “strategic” move to invest almost exclusively in a single type of supercomputing hardware, and the new government should instead focus resources on a range of supercomputing hardware, software and skills to match Britain’s AI and scientific needs.
Mark Wilkinson, director of the DiRAC facility and a theoretical astrophysicist at Leicester university, said it was sensible to pause briefly and reflect on the best ways to invest limited resources given that the “UK has fallen well behind comparable countries in terms of compute capacity”.
He cautioned that the government should not “salami-slice the funding into lots of pieces so everyone gets a small amount”, but focus instead on the best projects. The focus on “exaflops” was misplaced in the UK, he said, adding: “Flop count is only one measure of performance — what ultimately matters is productivity.”
Sir Peter Mathieson, principal of Edinburgh university — which has spent £31mn of its own funds on building its exascale supercomputer — has so far held back from public criticism of the government’s decision. He has met Kyle and said he welcomes “the continued dialogue on this vital issue”.
Still, several people have been highly critical of how Starmer’s government communicated its decision to rescind funding without promising any replacement.
Ben Johnson, former policy adviser to the last Tory tech secretary, said “unclear communication . . . can give rise to confusion in the community of researchers”.
“The market is extremely hot at the moment and is getting hotter by the week,” he added. “Any delay to procurement would potentially be very costly.”