Why AI Still Hasn’t Produced a World-Changing Drug

Biotech policy writer Ruxandra Teslo argues that the main obstacle to AI-led drug discovery is not model capability, but the system surrounding drug development. In her view, the biggest delays come from long clinical trial timelines, slow validation of surrogate endpoints, limited access to medical data, and a patent structure that rewards molecule design more than the discovery of new biological targets. She points to an average seven-year path from discovery to market and costs above $1 billion, saying much of that burden sits in regulatory process rather than science itself. Teslo also cites a 12-year validation path for bone mineral density as a surrogate endpoint in osteoporosis trials. Her broader argument is that AI can help optimize work after a target is chosen, but that is not where the industry’s deepest bottleneck sits. The incentive system still pushes companies and capital toward safer, patentable molecule design instead of riskier target discovery, leaving a core structural problem unresolved.