UK Sovereign AI fund backs Potential Sciences, which trains drug-design AI on physics simulations
The government says the Cambridge startup's eight-figure seed round includes Sovereign AI money and AIRR supercomputer access; the company claims its GPU quantum-chemistry code makes calculations 100 times cheaper.
Why it matters
Most chemistry AI learns from the same limited record of past experiments; Potential Sciences instead generates training data from physics-based quantum-chemistry simulations, which the government says could make design calculations 100 times cheaper. The deal is another use of the UK's £500 million Sovereign AI fund, and it comes with compute on the national AI Research Resource, including Isambard-AI in Bristol. The approach still has to prove itself against lab results. Watch for first drug-discovery partners and published benchmarks of the simulation data.
Key facts
- Announced by DSIT on 5 November 2026
- Eight-figure seed round including Sovereign AI
- Access to the AI Research Resource, including Isambard-AI
- AI model built on physics, not past experimental data
- Claims calculations 100 times cheaper
- Founder and CEO Dr Hatem Helal; based in Cambridge