Boltz $28M Seed Round: Open-Source AI Takes On AlphaFold 3

What happened

Boltz, the team behind a widely used family of open biomolecular AI models, has formally become a company. It launched as a public benefit corporation and announced three things at once: a $28 million seed round, a multi-year collaboration with Pfizer, and a preview of a hosted platform with two initial drug-discovery agents. 1 The round was led by Amplify Partners, Andreessen Horowitz, and Zetta Venture Partners. 12 Other backers include Hugging Face CEO Clement Delangue, Factorial Capital, and Obvious Ventures. 2

Gabriele Corso is CEO, with co-founders Jeremy Wohlwend and Saro Passaro. The company traces its technical roots to MIT CSAIL and Regina Barzilay's group. 1 The original Boltz-1 team also came out of MIT's Jameel Clinic and included Professor Tommi Jaakkola. 3 The company says its aim is to keep building open models and to package them so drug hunters can use them without in-house machine learning or infrastructure expertise. 1

The Pfizer piece

The Pfizer agreement is the most commercially telling part of the launch. Boltz will train exclusive models on Pfizer's proprietary data, covering structure prediction, small-molecule binding affinity, and biologics design. Its scientists will also work with Pfizer's discovery teams on custom models and workflows for specific target programs, with the goal of improving preclinical decision-making. 2 The collaboration, announced in January 2026, centers on Boltz's newer models, Boltz-2 and BoltzGen. 4

How Boltz got here

The company's credibility comes from a fast run of open releases. Boltz-1, released in November 2024, was presented as the first fully commercially usable model to reach AlphaFold 3-level accuracy on 3D structure prediction for biomolecular complexes. 2 It shipped under the MIT License, so anyone could use it for free, including for commercial work. 3 Later releases expanded its scope from structure prediction into affinity estimation and then generative design. 1

That path matters because of the benchmark Boltz was chasing. AlphaFold 3 appeared in May 2024 and marked a real generational jump. AlphaFold 2 handled proteins alone, while AlphaFold 3 could model complexes that combine proteins, nucleic acids, small-molecule ligands, and ions in one system. 3 Industry commentators describe the period since as a "post-AlphaFold" phase. In this view, AlphaFold 3's release had side effects that opened space for an open-source counter-movement led by Boltz and Chai. 3

Where the coverage agrees and diverges

Accounts of the launch agree on the core facts: the funding amount, the lead investors, and the Pfizer partnership. 124 The framing differs. Trade coverage emphasizes democratization and a bundled launch built around operational commitments. 12 Industry-analysis pieces instead place Boltz within a broader shift in foundation models. One treats it as part of a competitive open-source movement against AlphaFold 3. 3 Another presents the Pfizer deal as evidence of a pharma-wide pivot toward open biomolecular models. 4

Our read

The interesting tension here is not open versus closed in the abstract. It is how an open-weights company actually makes money. The Pfizer deal answers that question fairly clearly. The public models serve as the credibility engine and the funnel. The revenue comes from proprietary models trained on a partner's private data, plus a hosted product for teams that don't want to run infrastructure themselves. That resembles the open-core playbook familiar from software, applied to drug discovery.

This hybrid model is probably sensible, but it is worth watching closely. The value of open structure-prediction models rests on the idea that the community can inspect, reproduce, and extend them. If Boltz's best-performing capabilities increasingly sit inside exclusive, partner-specific models, the open releases could become a showcase rather than the frontier. The public benefit corporation structure, and the commitments Boltz made at launch, are presumably meant to reassure the research community on this point. 1 Whether they hold up will depend on what Boltz keeps releasing openly after the Pfizer work begins producing results.

For big pharma, the appeal is easy to see. A partner with a strong open track record lowers vendor lock-in risk while still turning internal data into a proprietary edge. For the wider field, a well-funded open challenger keeps pressure on AlphaFold 3 and on closed competitors to justify their restrictions. A $28 million seed is modest by AI-lab standards. Even so, it gives one of the most visible open efforts in structural biology the runway to test whether openness and commercial exclusivity can coexist.