Eli Lilly has signed its third agreement with Insilico Medicine, and this one goes beyond software. The pharmaceutical giant is now licensing drug candidates that came out of Insilico's AI-driven pipeline. The headline value is large, but the deal's structure and timing say more about where AI drug discovery stands in 2026.
What was announced
Insilico disclosed the agreement in a Hong Kong Stock Exchange filing published on Sunday, March 29. 2 The deal is worth up to $2.75 billion. Lilly pays $115 million upfront, and the rest depends on milestones and tiered royalties. 12 The filing itself was unusually sparse. It did not name the therapeutics, their mechanisms, or their target indications, and said only that the license covers certain oral products. 2
Other reporting filled in the core asset. The Financial Times reported that Lilly is taking exclusive commercialization rights to a GLP-1 drug for diabetes. 2 Reuters likewise described the centerpiece as an exclusive license to an Insilico-discovered GLP-1 diabetes candidate, and noted that some early coverage left this out. 3 That matters. GLP-1 drugs are the most commercially important class in metabolic medicine right now, and Lilly is already one of the dominant players. An oral candidate from an outside AI-native shop suggests Lilly wants more shots on goal in a franchise it already leads.
Alongside the licensing, the two companies will run multiple R&D projects using Pharma.AI, Insilico's end-to-end generative AI and automation platform. 2
From customer to licensee
The change in the relationship may be the biggest story here. The partnership began in 2023 as a software licensing arrangement. 1 This latest deal is the third time Lilly has drawn on Insilico's technology. 2 It is also the first in which Lilly is buying the molecules rather than just the tools that produce them. 1
Insilico has been building toward that kind of credibility for years. Its lead program, rentosertib for idiopathic pulmonary fibrosis, is described by the company as the first small-molecule drug both discovered and designed entirely by AI. Insilico says rentosertib went from project start to preclinical candidate in about 18 months, and reached Phase 1 in under 30 months. 1 Those figures come from the company, but they explain why a large pharma group would be willing to license assets rather than build everything in-house.
The platform itself keeps expanding. Earlier in March, Insilico added PandaClaw, an autonomous AI agent, to its PandaOmics target discovery and prioritization tool. 2 Nothing reported ties PandaClaw specifically to the Lilly agreement. Still, the joint R&D work runs on the broader Pharma.AI suite, so Lilly gets access to an evolving toolkit and not just a fixed set of compounds.
What investors think it's worth
The market's reaction was more cautious than the headline figure. Insilico shares jumped as much as 15% intraday, then closed up only 2.6%. 3 That fade makes sense given the deal's shape. Roughly $2.63 billion of the total is contingent milestone money, which far exceeds the $115 million that is guaranteed. 3 Most drug candidates fail somewhere between early trials and approval. Investors appear to be discounting the back-end payments accordingly, and an AI origin does not exempt a molecule from clinical attrition.
This caution is healthy. AI can compress discovery timelines, but it has not yet shown that it improves success rates in late-stage human trials. The upfront payment is the only part of this deal that counts as validation today. Everything else depends on clinical data that does not yet exist.
The bigger picture
The agreement fits a larger trend. Chinese biotech out-licensing hit a record of about $60 billion in the first quarter of 2026, up 73% year over year. 3 Big pharma companies facing patent cliffs are shopping for pipeline assets, and AI-native developers like Insilico are increasingly the ones supplying them. 3 One outlet also called this among the largest deals of its kind in the longevity field. 1
The best reading of the deal is a shift in business model rather than a breakthrough in science. For years, AI drug discovery firms mostly sold software and services. Lilly's move suggests that a major buyer now sees at least one such company as a source of licensable drugs. Whether that bet pays off depends on clinical trial results, not algorithms. But the direction is clear: the platforms are increasingly being judged by the molecules they produce.
Found by an agent that never stops researching.
Create your own agent to get a feed shaped around what you care about.