Last month, the UK government announced the home for its new exascale supercomputer, designed to give the country an edge in the global artificial intelligence race. The £900 million ($1.1 billion) project would be built in Bristol, a city in the west of England famed for its industrial heritage, and the machine itself would be named after the legendary local engineer, Isambard Kingdom Brunel.

The Brunel AI project should have been a big moment for another Bristolian export—Graphcore, one of the UK’s only large-scale chipmakers specializing in designing hardware for AI. Valued at $2.5 billion after its last funding round in 2020, the company is trying to offer an alternative to the US giant Nvidia, which dominates the market. With AI fast becoming an issue of geopolitical as well as commercial importance, and countries—including the UK—spending hundreds of millions of dollars on building strategic reserves of chips and investing in massive supercomputers, companies like Graphcore should be poised to benefit. In May, Graphcore’s CEO Nigel Toon wrote to the government, asking that some of the exascale project’s funding be allocated to British chipmakers—i.e., to his company.

But that deal hasn’t come through, and the company has struggled to turn early hype around its products into sales. This week, Graphcore filed accounts showing that it urgently needs to raise new funding. If it can’t do so by May next year, the company faces “material uncertainty” over whether it can remain a going concern, as losses mount.

“I think a lot of this [business] is really about being able to sustain your very capital-intensive development for long enough until you get acquired,” says Jakub Zavrel, founder and CEO of research company Zeta Alpha, which tracks the hardware used in AI research. “I think Graphcore has gotten squeezed in that game.”

Graphcore spokesperson Iain Mackenzie declined to comment on the company’s need to raise funding.

Founded in 2016 by Toon and Simon Knowles after the pair sold their previous hardware company to Nvidia, Graphcore has spent the last few years promising to build the next generation of chips. Instead of GPUs, graphics processing units, which are the current standard for AI applications, Graphcore focuses on IPUs, intelligence processing units. Graphcore claims its IPUs are better suited to the specific requirements of AI than GPUs, which are multipurpose chips originally designed for image processing. Early investors included Microsoft—now one of the giants in the vanguard of AI, and a big backer of OpenAI, developer of the ChatGPT chatbot. But in 2020, Microsoft stopped using Graphcore’s chips in its cloud computing centers.

Zavrel says that Graphcore may have struggled because its technology is significantly different from the Nvidia GPUs that users are familiar with. “I think what you see with Graphcore is that they are not able to take these researchers and engineers in a smooth way from the Nvidia-dominated ecosystem into their own thing, these IPUs that they’re producing,” he says.

The UK government’s current obsession with AI could have been an opportunity for Graphcore to get large-scale deals and to put itself in the shop window. Prime Minister Rishi Sunak has talked up his desire to turn the UK into a “technology superpower,” “the next Silicon Valley,” and the “home of AI.” Earlier this year, the government committed £1 billion to develop the domestic semiconductor industry, and £100 million to build a domestic reserve of chips, alongside hundreds of millions in other initiatives, including a “frontier models taskforce” looking into the risks and opportunities of advanced AI and a global summit on the existential threat of AI in November.

The initiatives have been criticized in some parts of the UK tech industry for excluding British companies, focusing on future risks rather than immediate opportunities, and for lacking ambition. The US and EU have committed tens of billions of dollars in subsidies for semiconductor manufacturing.

Being part of a supercomputer project used by academic and commercial researchers would give Graphcore visibility and mean more AI professionals were familiar with its technology.

Mackenzie, the Graphcore spokesperson, says that the company has been effectively cut out of the £100 million fund because the tender explicitly specifies GPUs, “thereby excluding systems built around Graphcore IPUs.”

“This is the realization of the warning that Graphcore issued in our open letter to the UK government—that a lack of technological diversity in our national AI compute infrastructure risks railroading users down the road of those applications that suit GPUs and limit exploration of models and techniques made possible by new, made-for-AI systems,” he says, adding that the US Department of Energy’s National Labs have made IPUs part of their infrastructure.

“Ironically, UK-based researchers can apply to use Graphcore IPUs via Argonne National Lab in the US,” Mackenzie says. “We would also reiterate that if the UK government is serious about nurturing an indigenous AI industry, it should consider that procurement is a powerful way of demonstrating that support—something we hope to see in future initiatives.”