Chai Discovery builds AI foundation models that design new drug molecules from scratch — antibodies, proteins, and small molecules — rather than screening libraries of existing compounds. Scientists input a disease target, and Chai's models generate candidate molecules tuned to bind it, before anything goes into a lab. The core products — Chai-1, Chai-2, and Chai-3 — are generative AI systems trained on the underlying physics and biology of molecular structure. Chai-2 was the first zero-shot platform for fully de novo antibody design to achieve double-digit experimental success rates. Chai-3 delivers tighter binding affinity against harder targets. Buyers are large pharmaceutical companies (Eli Lilly, Pfizer, Novartis, argenx) that access Chai's models via software agreements rather than buying drug assets. The pitch is deliberate: Chai is AI infrastructure for pharma — the same platform logic Databricks ran for data engineering — not a single-asset biotech betting on one drug.