Chai Discovery raises $400 million, valuing AI drug startup at $3.8 billion

Chai Discovery raised $400 million, giving the San Francisco AI drug-discovery startup a $3.8 billion valuation and sharpening investor bets that generative models can shorten the search for new medicines. The new financing lands as the company pushes from model benchmarks into partnerships with two of the world’s largest drugmakers.

The startup has moved fast since it was founded in 2024. In August 2025, Chai closed a $70 million Series A that brought its total funding to $100 million, and former Pfizer chief scientific officer Mikael Dolsten joined its board after that round. Its backers include OpenAI, Thrive Capital, Menlo Ventures, General Catalyst, Oak HC/FT and Dimension, a roster that reflects how aggressively venture capital has moved into AI-enabled biotech.

Chai’s strongest scientific claim rests on antibody design. The company has said its Chai-2 model produced a close to 20% hit rate in antibody design, compared with below 0.1% for traditional methods. Chai framed that result as a major step in de novo antibody design, where the goal is to generate workable candidates from scratch rather than sift through vast libraries of existing molecules.

Those results helped open doors with big pharma. On January 9, 2026, Chai announced a collaboration with Eli Lilly and Company to accelerate biologics discovery, and Joshua Meier, Chai’s co-founder and chief executive, said the effort was meant to bring important design capabilities to biotech companies in early discovery. In June 2026, Chai announced a license agreement with Pfizer that gave Pfizer scientists early access to Chai-3, the company’s most advanced model, along with a custom version trained on Pfizer’s proprietary data and tailored to Pfizer’s workflows.

The financing underscores a broader race among AI-biotech startups to turn computational promise into commercial tools that can survive the long timeline of drug development. Investors are rewarding milestone claims such as hit rates, platform access and licensing deals, but the harder proof will come later, when those models either help produce approved medicines or fade into another expensive experiment in drug discovery.