Seven months. That is how long it took Chai Discovery to nearly triple its valuation, from $1.3 billion in December to $3.8 billion this week. The AI drug discovery startup announced a $400 million Series C on Tuesday, and the timing was not an accident: it landed one day after the company disclosed a new collaboration with Novartis, on the heels of a June licensing deal with Pfizer and a January partnership with Eli Lilly.
Index Ventures led the round, joined by Kleiner Perkins, Sequoia Capital and Dimension. A wave of new backers, including Bain Capital Ventures, Battery Ventures, Baillie Gifford and Avra Capital, joined existing investors Thrive Capital, Oak HC/FT, Menlo Ventures and General Catalyst. OpenAI is also on the cap table, an unusual but increasingly common sight as foundation model companies quietly back startups applying their techniques to specialized domains like biology.
What Chai actually builds
Chai Discovery, founded two years ago, makes AI models that predict how biological molecules interact with each other. In practice, that means the company can suggest new antibodies and proteins optimized for a specific disease target before anyone touches a lab bench. Chai-3, the company’s most advanced model, is the product Pfizer licensed in June, alongside a custom version trained specifically on Pfizer’s internal data and workflows.
That pattern, licensing a general model plus a bespoke version trained on a pharma company’s proprietary data, has become Chai’s core business. Eli Lilly signed a similar arrangement in January. The Novartis deal, announced the day before the Series C, extends the same approach to antibody discovery specifically. For an industry that has spent billions on wet-lab trial and error, a model that can narrow the candidate pool before synthesis even starts is a genuinely attractive proposition, provided the predictions hold up once they reach a real laboratory.
Big Pharma’s AI dilemma
Pharmaceutical companies have a harder AI adoption problem than most industries. A hallucinated answer from a chatbot is an inconvenience. A wrong prediction about how a candidate antibody will behave in the human body is a wasted year and tens of millions of dollars, or worse, a safety risk that surfaces only in later-stage trials. That is part of why Chai’s traction with Lilly, Pfizer and Novartis carries weight beyond the funding numbers. Three of the largest drugmakers in the world do not sign multi-year AI licensing agreements as a marketing exercise.
It also explains why investors were willing to nearly triple Chai’s valuation in well under a year. The company’s total funding now sits around $630 million, and the jump from $1.3 billion to $3.8 billion reflects less a change in what Chai has built and more a change in how proven it looks. Commercial contracts with established pharma partners function as a kind of external validation that a research paper or benchmark score cannot provide.
A crowded, fast-moving field
Chai is not alone in this space. AI-driven drug discovery has drawn intense venture interest over the past two years, with rivals pursuing similar bets on protein structure prediction and molecular design. What separates the winners from the rest will likely come down to which models actually produce compounds that survive clinical trials, a process that takes years regardless of how fast the underlying AI improves. Chai’s Series C buys it time and credibility to keep working that problem, but the real test is still ahead, somewhere in a lab, months or years from now, when one of these AI-designed antibodies either works in a patient or it does not.
For more coverage of AI startups and funding, visit Mylistingo.
