Chai Discovery pursues $400M round for AI antibody design

An OpenAI-backed antibody design startup less than two years old is reportedly courting a $400 million round at a valuation in the low single-digit billions, which is a lot of money to raise on a claim that is still, mostly, a claim. That is the shape of what The New York Times' DealBook is reporting on Chai Discovery this week, and it is worth reading in the context of the last six months, not just today's number.

The backdrop matters. Chai closed a $130 million Series B in December 2025 at a $1.3 billion mark, and by early June Forbes reported the company was already in talks for another $400 million at roughly $3.4 billion, with no lead investor named. If the round closes anywhere near those terms, Chai will have tripled its valuation in about half a year on the back of Chai-3, a next-generation antibody design model it quietly deployed earlier this year and describes as a large step up on Chai-2, which itself was the first zero-shot antibody design system to hit double-digit experimental hit rates.

The commercial story pinned to that model is what pharma investors actually want to hear. Chai has a licensing agreement with Pfizer that includes early access to Chai-3 and a custom version trained on Pfizer's proprietary data, and a January partnership with Eli Lilly to build a purpose-built model trained exclusively on Lilly's internal discovery data. The compressed pitch is that an antibody design cycle that traditionally takes 12 to 24 months can be shortened to four to eight weeks. If that holds up outside curated benchmarks, it reprices a big chunk of the biologics workflow.

The honest caveat is that most of what makes the mark defensible is still prospective. As of the reporting we could confirm this run the round was 'in talks,' the lead had not been chosen, and the actual clinical output of the Pfizer and Lilly programs is not public. Foundation-model economics in biology have also not been tested against pharma's incentive to eventually train in-house on their own data. Take the specifics as reported, not settled.

What is worth watching is less the headline valuation and more whether Chai's second-tier pharma customers show up over the next twelve months. If mid-tier biotechs and non-US pharmas start licensing Chai-3, the 'AI infrastructure for drug design' framing starts to look real. If the customer list stays at two marquee names, it is a very expensive proof of concept.