Chai Discovery, an OpenAI-backed biotech startup now valued at $4B, is at the center of a shift in how pharma companies adopt AI tools. The post explains why pharma companies historically built their own drug pipelines instead of buying AI tools, and what changed: models became good enough to trust for drug design. Chai's binding models enable 'one-shotting' molecules — designing candidates that go straight to the clinic with minimal lab iteration — turning drug discovery into an engineering problem. Since June, Chai has closed deals with Lilly, Novartis, and argenx. The post previews a podcast with cofounder Matt McPartlon and product lead Neil Patil covering protein token value, abstraction levels in models, and how better tech reshapes pharma portfolios.
Questions this post answers
Why did AI drug discovery companies historically build their own drug pipelines instead of selling tools to pharma?
Convincing pharma to use an AI tool required proof it worked, which meant having validated drug targets. Once you had those, it was easier to raise money or sell a specific target than to license a tool across many pharma portfolios. The 'build your own drug' path was the only commercially viable route until AI models became good enough for pharma design teams to trust them directly. Developers and researchers tracking the AI-pharma commercialization shift follow deals like these on daily.dev.
What does Chai Discovery mean by 'one-shotting molecules' in drug design?
One-shotting molecules means designing drug candidates that are good enough out of the gate to proceed toward the clinic with minimal lab iteration. Chai's binding models predict how well a molecule binds to a target protein, enabling design rather than trial-and-error screening. This compresses iteration cycles and turns drug discovery into an engineering workflow where you hill-climb toward a final candidate computationally. Teams building at the AI-biology intersection track breakthroughs like this on daily.dev.