OpenAI is among the backers of a biotech founded in 2024 that is already valued at $1.3B.

TL;DR

Chai Discovery just raised a $130M Series B at a $1.3B valuation, backed by OpenAI, on the strength of a de novo antibody design model claiming a 100-fold improvement over previous computational methods. The bigger strategic question is whether the model layer commoditizes — and if so, where value actually migrates. I see a fascinating market split forming between molecule design and target identification, and the real prize is integrating both.

OpenAI is among the backers of a biotech founded in 2024 that is already valued at $1.3B.

Chai Discovery raised a $130M Series B at a $1.3B valuation, four months after a $70M Series A, bringing total funding to over $225M.

The pitch is de novo antibody design.

Their Chai-2 model claims 16% success rates in designing functional antibodies from scratch, a 100-fold improvement over previous computational methods.

That sounds impressive until you remember what comes next.

Designing an antibody that binds is the first step. Getting it through manufacturability, stability, immunogenicity, and eventually clinical trials is where most programs die. The gap between "binds in-silico" and "works in patients" has swallowed billions in pharma R&D.

Chai knows this. Their recent preprint emphasizes "developability" properties, claiming 86% of designed antibodies meet preclinical selection standards. They are trying to close the gap before pharma dismisses this as another pretty model.

But here's the strategic question: is antibody design becoming infrastructure, or a moat?

AlphaFold opened structure prediction. Isomorphic Labs, Recursion, and now Chai are all racing toward generative molecular design. If the model layer commoditizes, value migrates to proprietary data, wet lab validation, and clinical execution.

This creates a fascinating split in the market.

Last week, I wrote about pharma paying earlier for target identification (Novartis x Relation). With Chai, investors are pricing the molecule design layer itself.

That said, the real prize is not one or the other, It is integrating both.

If Chai's 16% holds up in the real world, discovery starts to look less like a cost center and more like a compile step. Not because biology is "solved," but because iteration gets dramatically cheaper.

It is clearly an exciting time to be in TechBio in 2025.

Key takeaways

• Chai Discovery raised $130M Series B at a $1.3B valuation just four months after its Series A, with OpenAI among the backers — a signal of how fast capital is moving into AI-native biotech.
• Their Chai-2 model claims 16% success rates in de novo antibody design, which they frame as a 100-fold improvement over previous computational methods.
• The gap between "binds in-silico" and "works in patients" has swallowed billions in pharma R&D — designing an antibody that binds is only the first step.
• Chai is trying to pre-empt pharma skepticism by emphasizing developability, claiming 86% of designed antibodies meet preclinical selection standards.
• If the model layer commoditizes — as AlphaFold did for structure prediction — value will migrate to proprietary data, wet lab validation, and clinical execution.
• I see a market split forming: pharma is paying earlier for target identification (Novartis x Relation) while investors are now pricing the molecule design layer itself with Chai.
• The real prize is not one or the other — it is integrating both target identification and molecule design into a single, faster discovery engine.