The AI Drug Discovery Revolution

AI could transform drug discovery from a costly, uncertain trial-and-error process into a more systematic engineering discipline, improving success rates and making more diseases treatable.

• Bringing one therapy to market costs an estimated $2 billion, only slightly more than 1 in 10 drugs entering Phase I reaches patients, and roughly 21,000 drugs are currently in development.
• AI is already automating documentation-heavy clinical workflows that can take up to 14 weeks, while future applications could improve trial design and identify safety risks earlier.
• Drug developers remain concentrated on a small number of validated targets: 37 targets each have more than 50 competing drug programs, highlighting the need to uncover new disease mechanisms and therapeutic opportunities.
• Insilico Medicine's rentosertib, whose target and molecule were designed with generative AI, has produced positive Phase IIa results, offering early evidence that AI can generate differentiated medicines.
• Biological AI models are expanding rapidly, from single-digit annual launches to more than 380 per year, but future foundation models may require thousands of times more data than many current biological datasets.

Takeaway: Investors should view AI as a potential driver of a healthcare R&D productivity revolution over the next decade, although the opportunity remains dependent on better biological data and higher clinical success rates; Tema presents its Healthcare AI ETF as one way to access companies such as Revolution Medicines and Vertex Pharmaceuticals.