Sovereign AI backs Cambridge chemistry startup Potential Sciences with cash and supercomputer access

The UK government's Sovereign AI venture fund has backed Potential Sciences, a Cambridge startup building AI models for chemistry, as part of what the government calls "an eight-figure fundraise". The investment, and access to national supercomputing, were announced in a gov.uk press release published on 5 November 2026 by the Cabinet Office and the Department for Business, Innovation, Science and Trade.

The round

• Size: the government describes it as "an eight-figure fundraise". The company's own launch post says: "We have raised eight figures in capital and compute", which on our reading counts the compute as well as the cash. Neither gives a figure. WIRED reports that the founders "declined to disclose the precise figure or valuation".
• Who invested: neither the government nor the company names a lead investor. Potential Sciences calls the Sovereign AI fund and Twin Path Ventures its "key investors". The government says other backers include Twin Path Ventures, Odyssey, Portfolio Ventures, Blue Wire Capital, rule30, Nauta Capital and Lightfield Equity, as well as angel investment from the likes of Patrick Collison (co-founder of Stripe) and Chris Gibson (co-founder and former CEO of Recursion).
• Sovereign AI's cheque: not stated. The release's notes to editors say: "Like most venture capital funds we do not comment on specific investment figures as they are commercially sensitive."
• Compute: the government says Sovereign AI "is also making compute available to the company through the AI Research Resource, the UK's national suite of advanced supercomputers, including the Isambard-AI supercomputer in Bristol." It gives no number of GPU hours or chips.
• Advisers: the government says AI engineer Andrew Fitzgibbon and entrepreneur James Hawkins will advise the company.

WIRED also reports that Joséphine Kant, whom the government names as a Sovereign AI Investment Partner, has joined the startup's board as part of the investment.

What the company says it does

In its launch post, Potential Sciences describes itself as "a frontier lab for chemistry". It says: "We generate new training data from the laws of physics, at the accuracy of a laboratory measurement."

The company's case is that AI models in chemistry are short of data, because they learn from a limited record of past lab experiments, or from simulations it calls "too crude to trust". Its answer is to compute the data instead. The post says its "GPU-native quantum chemistry engine" runs "gold-standard CCSD(T) calculations at 100x lower cost than traditional implementations". It plans to use that data to train what it calls "atomistic foundation models".

The company also says its approach has let it "beat the published state of the art" on predicting reaction barriers and protein-ligand binding affinity. It has not yet published those results: "We will share full results in a coming release." It says it will start with small-molecule drug discovery, then move to "catalysts, enzymes and materials".

According to the post, Hatem Helal founded the company in October 2025. It is based in Cambridge, with researchers in Bristol and Tübingen.

What the government says Sovereign AI is

The release calls Sovereign AI "the government's big bet on promising early-stage AI companies to help them grow, scale and succeed from Britain". The notes to editors say the fund "will invest directly in promising UK AI companies to help them scale in the UK", focused on early and growth-stage companies, and that "Typical equity investments will be worth around £1-10 million".

The release does not give the fund's total size. It does say the news "comes a week after the Budget, where the Chancellor confirmed a further £140 million backing for Sovereign AI", made up of £60 million for the fund's investments and £80 million for its R&D Procurement Scheme. It adds that Sovereign AI will expand into "Physical and Industrial AI".

AI Minister Kanishka Narayan said solving chemistry's "data bottleneck" could let researchers "explore new treatments for diseases that are currently deemed too rare or expensive to focus on".

Why it matters

On our reading, the deal shows the shape Sovereign AI's support is taking: a stake plus access to state-funded supercomputers, for a company whose product depends on raw compute. The scale of both is undisclosed, so how much each one is worth cannot be judged from what has been published. The company's performance claims are its own and are not yet backed by published results.