Patrick Collison, the Stripe chief executive, and the UK government’s Sovereign AI fund have both put money behind Potential Sciences, a London company only months old that wants to model chemistry from first principles rather than from lab notebooks.
Executives at the startup will not say what the raise came to, beyond calling it tens of millions of dollars, and they are equally quiet about valuation.
Collison put in money as an angel, with others, and venture money arrived through Twin Path Ventures, Odyssey, Portfolio Ventures, Blue Wire Capital, rule30, Nauta Capital and Lightfield Equity.
Training on laws, not lab results
Where most computational chemistry tools learn from experimental datasets, this one leans on the fundamental laws of physics, combined with quantum chemistry simulation, to predict molecular behaviour before any bench work happens.
That approach, the company argues, could make the computation behind chemical design 100 times cheaper. Should it hold up, the real bottleneck in drug discovery moves away from computing and toward proving molecules in a wet lab.
Capital is not the only thing the fund is offering. It is handing over compute capacity too, a structure governments increasingly favour as they try to keep frontier labs anchored at home.
A founder from the chip wars
Hatem Helal, a medical doctor who once worked as an executive at Graphcore, started the business in November 2025. The Bristol chip designer he left behind was bought by SoftBank in 2025.
Olix, a chip startup, sits in the same portfolio, as do AI outfits Ineffable Intelligence and Callosum, which suggests a fund chasing the compute and science layers rather than consumer apps.
Chemistry is among the hardest tests AI faces. Bold promises about cheaper simulation only matter if the molecules they predict turn out to work in practice.
