London has opened a $675 million fund to back homegrown AI companies.

Fund structure and incentives

The Sovereign AI fund will deploy roughly $675 million to British startups working on core AI technology, ranging from model development to agentic systems and drug discovery. Alongside cash, the fund gives startups access to the UK's supercomputer fleet, easier visas for overseas hires, routes into government procurement, and hands-on advice from civil servants and specialists.

James Wise, partner at Balterdon Capital, will co-lead Sovereign AI. Joséphine Kant, formerly of Dogwood Ventures and Y Combinator, joins Wise as the fund’s other lead. Both leaders come from venture firms rather than government, and they'll run the fund's investment decisions and work directly with portfolio companies.

The government framed the fund as a way to shift the UK from being mainly a user of foreign AI tools to producing more of its own. Liz Kendall, the United Kingdom’s technology secretary, said the initiative is “unlike anything Government has ever done before” and argued the approach will remove barriers that have restrained British firms.

Kendall added that the effort will help secure Britain’s economic prosperity and national security in the modern age.

Early bets and compute grants

The fund has already made a visible early investment in Callosum, a startup building software to coordinate different classes of processors so they can work together effectively. That kind of work matters as companies try to pair GPUs, CPUs, and specialized accelerators to squeeze more performance out of expensive hardware.

In addition to direct equity, Sovereign AI awarded up to 1 million GPU hours apiece on the UK’s supercomputer network to six startups: Prima Mente, Cosine, Cursive, Doubleword, Twig Bio, and Odyssey. The recipients will use the compute time to train models and run simulations that would be costly or slow on private infrastructure.

The compute grants cut what startups must spend to run experiments. For VCs, the scheme reduces technical risk on costly experiments while leaving the equity upside and future sales untouched.

Where the UK starts from

The UK already hosts prominent AI firms and research centers. Google DeepMind, ARM, and autonomous driving firm Wayve are based in Britain and have produced high-profile breakthroughs and talent. Still, parts of the AI value chain remain concentrated in the United States and Asia, especially semiconductor design and large-scale model training.

Officials say backing domestic firms could help the UK grab a larger share of global AI investment and cut reliance on foreign tech that might be used as leverage in trade or security talks.

Rosaria Taddeo, professor of digital ethics and defense technologies at the University of Oxford, told Wired earlier that the prevailing idea that innovation is only happening in the U.S. Has been damaging. “We have been too gullible to the narrative that innovation is done in the US—that we lost the AI train and shouldn't even think about it,” Taddeo said. “That’s a dangerous narrative.”

Limits and risks

Experts caution the UK can't expect total self-sufficiency in AI. General-purpose model development remains dominated by U.S.-based firms like OpenAI, Anthropic, and Google, which command vast compute, talent pools, and distribution channels. The UK’s fund doesn't promise to replicate those exact capabilities.

Analysts warn that cutting off international partners would probably slow progress and leave UK firms facing higher costs or weaker performance. Instead, the fund is being pitched as a way to strengthen homegrown capability in areas where the UK can realistically compete or carve out a niche—software that optimizes heterogeneous hardware, specialized agents, and biotech applications tied to simulation and modeling.

Even with government support, building up semiconductor fabrication or giant-scale model training in the UK would require time, sustained capital, and supply-chain alignment. Policymakers will have to weigh how much public money to commit, how to attract private co-investment, and how to structure interventions without crowding out venture investors.

Investor and market implications

For venture investors, a government-backed fund that combines equity with nonfinancial support is a new signal. It suggests the UK will deploy a mix of carrots—compute, visas, procurement pathways—and sticks—preferential treatment for domestic suppliers in government contracts—to cultivate local champions.

That could tilt deal flow toward startups that can show clear national or industrial value, such as companies working on defense-adjacent tools, drug discovery, or infrastructure that eases reliance on foreign chipmakers. At the same time, the fund’s involvement may encourage co-investment from private VCs seeking government-backed de-risking on expensive technical milestones.

Institutional investors will want clear governance — how the fund values and prices deals, its expected exit timelines, and how it mixes commercial returns with strategic goals. Those details will shape whether the fund becomes a scalable lever for private capital or a small, targeted program with limited market impact.

Policy context and the 2025 plan

Sovereign AI is one piece of a broader UK strategy first laid out in January 2025 to “position the UK to be an AI maker, not an AI taker.” The plan includes measures across research funding, talent policy, procurement, and regulation intended to foster domestic capacity while managing risks tied to advanced AI systems.

Tying compute and hiring help to funding is intentional and meant to lower barriers for founders. Large-scale model training and algorithm research are compute- and talent-intensive. By bundling those resources, the government aims to lower the initial hurdle that stops small firms from tackling bigger technical problems.

Still, how the program interacts with international collaboration will matter. The UK’s industry strength has long rested on global research networks and cross-border talent flows.

Maintaining open channels while scaling domestic capability is the balancing act policymakers face.

What to watch next

Investors and startups will be watching which sectors the fund prioritizes next and how quickly it deploys capital beyond its initial recipients. The fund’s co-leads, James Wise and Joséphine Kant, will be central to those choices, deciding which startups receive follow-on funding and which partnerships with industry and academia are worth backing.

Markets will also look for clarity on how compute grants are allocated and whether the program expands access to private-sector compute partners. The answers will show whether the initiative can move beyond symbolic support to materially change the economics of building advanced AI in the UK.

Related Articles

Sovereign AI has awarded up to 1 million GPU hours each to Prima Mente, Cosine, Cursive, Doubleword, Twig Bio and Odyssey.