Pharma stops building AI pipelines, buys Chai Discovery instead

The more consequential detail buried in the open-source-to-enterprise pivot is what Chai Discovery is giving up: community-driven model improvement and the credibility that came with it. Closing proprietary deals with Lilly, Pfizer, and Novartis means future model performance claims will be harder to independently verify, which is precisely the kind of trust problem pharma said it needed solved before adopting outside AI infrastructure.

This is largely disconnected from recent activity in our archive, as we have no prior coverage of Chai Discovery or AI-native drug discovery tooling to anchor against. That gap is itself worth noting: the biology-as-software thesis has been building for several years across protein folding, generative chemistry, and now structural biology workflows, but Modelwire has not yet established a thread on this vertical. The Chai story would be a reasonable place to start one, particularly given that pharma enterprise adoption is now a documented data point rather than a projection.

Watch whether any of the three named pharma partners (Lilly, Pfizer, Novartis) publicly discloses a Chai-assisted compound reaching IND-enabling studies within 18 months. That would be the first externally verifiable signal that the trust threshold crossed in the sales cycle is also holding up in actual research pipelines.