Chai Discovery

Chai Discovery has posted the single most striking benchmark in AI biologics: Chai-2, its zero-shot de novo antibody design model, reports a near-20% experimental hit rate against targets — a step change from the roughly 0.1% success of traditional library screening and prior computational methods, and the first such platform to reach double-digit hit rates while producing full-length, drug-like monoclonal antibodies rather than fragments. That collapses discovery cycles from months to weeks. The company pairs a genuine open-science land-grab — Chai-1, an AlphaFold-class structure-prediction model released openly, seeding the field with Chai as default infrastructure — with proprietary therapeutic-design capability on top. The founding team is elite (CEO Joshua Meier came from Absci, Meta AI, and OpenAI; co-founders from Stripe and structural-biology research), and the capital reflects it: a $70M Series A in 2025 with OpenAI, Thrive, Menlo, and Dimension, then a $130M Series B in December 2025 co-led by Oak HC/FT and General Catalyst at a $1.3B valuation, taking total funding near $230M. The January 2026 Eli Lilly collaboration is the commercial proof point — Chai builds a custom AI model exclusively for Lilly, trained on Lilly's proprietary data, validating the platform against the toughest possible customer. If de novo antibody design is real, Chai is the front-runner in the highest-value application of generative biology.

A $1.3B valuation on a company founded in March 2024, with roughly 11-50 employees and no clinical asset of its own, prices in near-flawless execution of a thesis that remains a benchmark rather than a drug. The near-20% hit rate is company-reported against selected targets; the industry has repeatedly watched in-silico design metrics fail to survive developability, manufacturability, immunogenicity, and — the only test that matters — the clinic, where AI-designed biologics have yet to produce an approval. Chai's open-model strategy is a double-edged sword: releasing Chai-1 openly builds mindshare but hands the structure-prediction layer to competitors for free, and the antibody-design frontier is fiercely contested by Generate Biomedicines, Absci, Xaira, EvolutionaryScale, and the internal teams at every large pharma, several of which have far more proprietary experimental data. The Lilly deal is a services-and-milestones collaboration, not recurring product revenue, and pharma partnerships of this shape are easy to sign and easy to quietly wind down. The business model is unresolved — is Chai a tools vendor, a services shop, or an eventual drug developer? — and each path has different, unproven economics. OpenAI's involvement generates halo but also dependency risk on a backer with its own shifting priorities. For a two-year-old company, the gap between a headline hit rate and a durable, defensible franchise is the entire risk.