THE ESSENTIALS IN 30 SECONDS

  • Potential Sciences, founded in London in November 2025 by former Graphcore executive Hatem Helal, has raised a seed round reported in the tens of millions of dollars.
  • The UK's Sovereign AI fund joined venture firms including Twin Path Ventures and Nauta Capital, plus angel investor Patrick Collison, and is also providing compute.
  • The company says its physics-based approach can make the AI calculations behind chemical design about 100 times cheaper; no model results have been published.

Potential Sciences, a London startup that wants to supply AI drug discovery with chemistry data computed from physics rather than measured in the lab, has raised a seed round backed by the UK government's Sovereign AI fund. The backing was confirmed on November 5. Neither the company nor the fund disclosed the amount, but Tech.eu describes the raise as being in the tens of millions of dollars.

Who is investing

Alongside Sovereign AI, the round includes Twin Path Ventures, Odyssey, Portfolio Ventures, Blue Wire Capital, rule30, Nauta Capital and Lightfield Equity, according to Tech.eu. Stripe chief executive Patrick Collison is among the angel investors, and Resultsense also lists Recursion co-founder Chris Gibson. The size of the state fund's stake was not given; Resultsense notes that its cheques typically range from about £1 million to £10 million.

The government package combines money with computing power. Resultsense reports that the company will get time on national supercomputers through the AI Research Resource, including Isambard-AI in Bristol. Sovereign AI has previously backed chip startup Olix and the AI companies Ineffable Intelligence and Callosum, Tech.eu says.

Physics-based data for drug design

The company was founded in November 2025 by Hatem Helal, a former Graphcore executive, who leads it as chief executive. Most AI models for drug discovery learn from experimental results, which are slow and expensive to produce. Potential Sciences instead trains its models on data generated by simulating molecules from the fundamental laws of physics. Helal says the team has rewritten quantum chemistry methods to run on GPUs, so that such data can be produced cheaply and at scale. RuntimeWire describes the goal as building a frontier lab for chemistry and credits Helal with a quantum chemistry dataset of about a billion examples.

What remains unproven

The company claims its approach makes the calculations needed to design chemicals roughly 100 times cheaper. That figure comes from the founder, and no model results or benchmarks have been published so far. Whether simulated data can stand in for laboratory measurements well enough to produce useful drug candidates is the central question the startup will need to answer.