Big Pharma's AI bet comes with fine print

TL;DR: Eli Lilly has signed a deal worth up to $2.75 billion with Insilico Medicine, a startup that uses AI to develop drugs. Only $115 million of that is upfront—the rest is contingent on milestones that could take many years to hit. It's the latest move in pharma's all-in bet on AI, which has proven it can speed up early-stage drug development. But no completely AI-discovered drug has cleared FDA approval yet, and the hard part is still ahead.

What happened: Lilly gets exclusive worldwide rights to a specific portfolio of Insilico's preclinical AI-designed drugs. This is the two companies' third deal in three years: a software license in 2023, a $100 million research collaboration in late 2025, now this. This latest agreement fits a deliberate broader AI strategy for Lilly—including a co-innovation AI lab with Nvidia announced in January, with up to $1 billion invested over five years. It's also part of a wider push across Big Pharma, which is reportedly staring down an estimated $300 billion in prescription drug revenue exposed to patent expiration by 2030 and needs to reload pipelines fast.

What AI has proven: The early-stage speed gains are real. Insilico's rentosertib—the first drug where both target and molecule were identified entirely by generative AI—went from concept to clinical candidate in 18 months (versus the standard three to six years through traditional methods, according to Insilico). The core case for AI drug discovery may not be about drugs reaching patients faster—it's about more shots on goal at a fraction of the per-shot cost. If you can generate candidates faster and cheaper, even an unchanged clinical failure rate should eventually produce more approvals.

What’s unproven: AI hasn't yet demonstrated it can improve on the roughly 90% clinical trial failure rate for drug candidates—and trials are where the real money burns. A drug can take around 10–15 years to reach market from discovery and costs anywhere between $1.1 billion and $2.6 billion. Right now, AI doesn't widely touch the back half of the process (but some GLP-1 makers have started to integrate AI into trials). Lilly's own CFO, Lucas Montarce, said in March that the company is "investing heavily" in AI for R&D, but that it will "take more time" to get AI drugs from research to clinical testing, according to the Financial Times.

Bottom line: Whether the shots-on-goal math holds is a question 2026 will start to answer, as the first wave of AI-designed candidates enters Phase 3 clinical trials—including for asthma and pulmonary fibrosis. The Eli Lilly-Insilico deal's own structure captures where things stand: $115 million in hand, $2.6 billion in hope. —SM