The world of drug discovery has just witnessed what might be its next great leap: Chai Discovery, a young San Francisco startup led by a team of AI veterans and bioengineers, has snagged $70 million in Series A funding with a mission to transform how we invent medicines. Backed by leading names like Menlo Ventures, OpenAI, and Thrive Capital, and now further bolstered by the addition of pharma legend Mikael Dolsten (ex-Pfizer CSO) to its board, Chai Discovery is positioning itself at the bleeding edge of AI-driven biology.
What Makes Chai Discovery Different?
The traditional hunt for new antibodies—a crucial building block for many drugs—has always been a slog: scientists must sift through millions, sometimes even billions, of possibilities just to find one that actually works. Even the best computational approaches typically yield a measly 0.1% success rate.
Enter Chai-2: the company’s new “all-atom generative model.” This AI system doesn’t just optimize existing blueprints; it generates entirely new antibody designs based on only a target and epitope—producing solutions that the lab can validate in days, not years. The numbers are eye-opening: In recent lab tests, Chai-2 hit nearly a 20% success rate across roughly 50 antibody targets, slashing both timelines and costs.
How Is Chai’s AI Model Chai-2 Changing Antibody Discovery?
Chai-2 isn’t just an incremental improvement. By taking a zero-shot approach—making all initial determinations solely from first principles and structure, rather than reusing prior biological knowledge—it advances the field from the “needle-in-a-haystack” search of classic methods to something more like “directed design.”
Speed: A process that took years and millions of dollars can now yield working antibodies in as little as two weeks.
Versatility: Chai-2 supports formats including nanobodies and miniproteins—crucial for next-generation therapeutics.
Robustness: Lab results show a 15-20% “hit rate” for new, previously untargeted antigens—a more than 100x improvement over older computational approaches.
Minimal screening: Chai-2 generates a shortlist of high-potential candidates, eliminating massive expensive screens.
The upshot? Drug developers now have a shot at previously “undruggable” or prohibitively expensive targets.
What’s Next, and Why Does It Matter for Drug Development?
This substantial funding round, and the prestigious addition of Mikael Dolsten (whose leadership at Pfizer produced over 36 approved drugs), reflect a giant vote of confidence from pharma and tech investors alike. Chai Discovery, founded by Joshua Meier (ex-OpenAI, Facebook AI), Jack Dent (ex-Stripe), and an A-list team of AI researchers, now plans to open its platform—on a selective basis—to pharmaceutical partners.
The aim: Make drug discovery less about trial-and-error and more about rational engineering. With Chai-2, partners could target diseases long considered immune to smart therapies, accelerate vaccine projects, or quickly address emerging threats with bespoke biologics.
Industry observers are taking note, with Menlo Ventures and Anthropic’s Anthology Fund lauding Chai’s combination of technical depth and real-world impact. If their platform scales, it could rewrite the rulebook for an industry desperately in need of speed and precision.
Additional Highlights
Chai-2 found a solution in hours to one antibody problem that had cost another company over $5 million and three years, with validation happening in under two weeks.
Using only a target antigen and epitope, Chai-2 can generate—in a “Photoshop for proteins” style—designs tailored to exactly fit their biological target.
Antibody drugs, vaccines, and emerging therapies could become dramatically faster and cheaper to develop, bringing new hope for diseases that have stymied researchers for decades.
Chai Discovery’s commercial rollout will be selective, focusing on responsible deployment and vetted partnerships.
Chai Discovery’s rise illustrates just how far artificial intelligence (AI) has come in reshaping the basic science of drug creation. With a platform that can iterate on designs as easily as a digital artist tweaks an image, the future of biotech innovation suddenly looks a lot more programmable—and potentially, a whole lot faster.
