Chai Discovery
NativePrivateUpdated November 3, 2026AI molecular design/drug-discovery platform
- Website
- www.chaidiscovery.com
- CEO
- Joshua Meier
- Founded
- 2024
- HQ
- San Francisco, United States
- Total Funding
- $630M
- Employees
- 11-50
Products & Offerings
Chai-3
Chai's latest generative model for designing antibodies, which the company says roughly doubles Chai-2's target success rate and improves binding affinity.
Chai Platform (computer-aided design suite for molecules)
Licensed software that puts Chai's models in the hands of pharma scientists for de novo antibody and biologics discovery.
Bespoke partner models
Custom versions of Chai's models trained on a partner's proprietary data, such as the exclusive model built for Eli Lilly and the dedicated model for Pfizer.
Chai-2
Zero-shot generative model that designs antibodies from a target structure alone, reaching double-digit experimental hit rates across 52 targets.
Chai-1
Open-source model that predicts the structures of proteins, small molecules, DNA, RNA and covalent modifications, widely used and cited by academic researchers.
Executive Summary
What they do
- Builds generative AI foundation models that design antibodies from scratch (de novo)
- Licenses Chai-3 and custom models trained on each partner's own data to pharma
- Open-sourced Chai-1, a structure prediction model for proteins, small molecules, DNA and RNA
Why it matters
- Chai-2 hit about 16-20% experimental binding rates designing antibodies from target structure alone
- Says Chai-3 roughly doubles Chai-2's target success rate, with tighter binding
- Signed Lilly, Pfizer, Novartis, argenx, BMS and GSK within about nine months
Best fit
- Large pharma biologics groups looking to shorten antibody discovery from months to weeks
- R&D teams with proprietary antibody data who want an exclusive, custom-trained model
- Programs chasing hard targets such as GPCRs or multi-specific antibodies
Watch-outs
- No Chai-designed antibody has reached clinical trials, so real-world value is unproven
- Small team of about 40-50 people serving many large pharma partners at once
- Reported deals run to mid-eight figures a year, out of reach for smaller biotechs
Latest News
AI Magazine · 4 days ago
PositiveContract Pharma · 25 days ago
PositiveFierce Biotech · 25 days ago
PositiveBusiness Wire · 25 days ago
PositiveUnlock Full Intelligence
Pro members get the complete briefing
- Complete analyst verdict
- Risk assessment
- All 6 Bluescore dimension scores
- Traction signals
- Bluescore history & momentum tracking
- Unlimited company access
Chai Discovery is the most commercially validated startup in AI antibody design. In 2026 alone it signed platform deals with Eli Lilly, Pfizer, Novartis, argenx, Bristol Myers Squibb and GSK, and raised a $400M Series C at a $3.8B valuation. Pharma and large biotech biologics teams should run a structured pilot on their own targets and data, comparing hit rates, developability and IP terms against in-house methods and rivals such as Generate Biomedicines, Absci and Isomorphic Labs. Proof that its designs succeed in the clinic is still missing.
Signal Provenance
Latest observation date and source for each signal category
| Signal Category | Last Updated | Source |
|---|---|---|
Funding Signals Last funding round, Funding stage | Oct 31, 2026 | Montclare research agents |
Hiring Velocity Headcount, Headcount growth | Nov 3, 2026 | Montclare research agents |
Technical Output Research papers, GitHub activity | Nov 3, 2026 | Montclare research agents |
News Sentiment News mentions, News sentiment | Nov 1, 2026 | Montclare research agents |
Partnership Signals Enterprise partnerships, Distribution partners | Nov 3, 2026 | Montclare research agents |
Revenue Indicators ARR estimate, Revenue growth | Nov 3, 2026 | Montclare research agents |
Leadership Signals CEO tenure, Founder-led | Nov 3, 2026 | Montclare research agents |
Competitive Position Category rank, Market share | Never | Montclare research agents |
Why this Bluescore?
6 dimensions, each scored 0–10 against all — companies. Each adds its weight × score to the Bluescore — see methodology