Overview
- The UK’s Sovereign AI fund invested in Potential Sciences and granted access to the AI Research Resource including the Isambard-AI supercomputer, a package announced on Thursday, November 5.
- Potential Sciences says it has rewritten quantum-chemistry code to run on GPUs and can make the calculations needed for chemical design up to 100 times cheaper, enabling large-scale synthetic data generation for model training.
- The startup’s physics-first approach seeks to simulate atomic and molecular interactions to train predictors that can rule out poor drug candidates early and make rare-disease drug searches more economical.
- Experts and the company caution that faster in-silico candidate selection does not remove the long, costly steps of preclinical and clinical safety and efficacy testing that determine whether a drug reaches patients.
- Key uncertainties remain over technical validation at scale, the company’s commercial and openness plans, and governance of dual-use risks as Potential Sciences moves from early funding toward product and regulatory milestones.