Pharma stopped building its own AI and started licensing the tools

Drug design software just stopped being something pharma companies build in-house and started being something they license. For years, every AI-for-pharma startup was pushed toward running its own drug programs because pharma would not trust the tools. That hedging has now ended. Latent Space's interview with Chai Discovery's founders puts a stake in when: when the models got precise enough to design bi-specific antibodies and trigger biological cascades that resist hand design, in-house AI teams stopped being the safer bet. Discovery groups now buy the software and keep the candidate-selection work inside their own walls.

A two-year-old company at the center of four AI-pharma tools deals in one January week is the market's way of pricing that switch. Latent Space frames four deals in one J.P. Morgan Healthcare Conference week as the visible edge of a category-wide repricing — and that count is Latent Space's, not an independently enumerated market tally — anchored by Chai Discovery's $3.8B Series C valuation per LinkedIn and IntuitionLabs, with a post-deal valuation reaching $4B per Latent Space. The mechanism will repeat wherever applied AI crosses a trust line: founders stop being told to verticalize into their customer's product, and buyers start signing tools contracts instead. For drug pipelines, the consequence is more candidates reaching the clinic, including bi-specific and cascade-triggering antibodies that wet-lab discovery alone could not reliably reach.