Boltz, an AI drug-design lab spun out of MIT and now based in London's King's Cross, has raised a 110 million dollar Series A led by Bessemer Venture Partners. The round, reported by Tech.eu on November 19 and confirmed by Bessemer's announcement, brings back all existing investors, including Andreessen Horowitz, Amplify Partners, Zetta Venture Partners and Obvious Ventures, and adds Greycroft, F-Prime Capital, Hummingbird and Benchling as new participants. No valuation was disclosed. It is nearly four times the 28 million dollar seed round the company announced in January.
Boltz sells what it calls programmable molecular design the way frontier labs sell language models: by the token. Its commercial models, BoltzMol-1 for small molecules and BoltzProt-1 for proteins, launched in June behind a consumption-priced API that also plugs into coding agents such as Claude Code, Codex and Gemini. A chemist uses the models to score how strongly candidate molecules bind to a target protein and to generate new molecular designs, reviewing a shortlist instead of synthesizing and testing hundreds of thousands of compounds.
The traction numbers, which come from the company and its investors rather than an independent audit, are the reason a Series A of this size exists: Boltz says it has signed partnerships with six of the world's 20 largest pharmaceutical companies, naming Pfizer, GSK, Takeda and Bristol Myers Squibb, and that molecular token usage among paying customers has grown roughly 50-fold since June. Bessemer adds DSM Firmenich to the partner list.
The company's lineage explains the adoption. Founders Gabriele Corso, now CEO, Jeremy Wohlwend, now CTO, and research head Saro Passaro worked in Regina Barzilay's lab at MIT, where they released Boltz-1, an open-source biomolecular structure-prediction model that aimed to match DeepMind's AlphaFold3. Corso previously contributed to DiffDock, an early generative-AI approach to molecular docking. Bessemer says the open-source models have been downloaded more than 5 million times and used by more than 100,000 scientists, and Nvidia invested earlier in 2026 during Jensen Huang's visit to the UK, according to Tech.eu. The team has grown to about 40 people.
Performance claims remain vendor-reported. Boltz says BoltzMol-1 found small-molecule hits on 6 of 10 hard targets in an internal test and that Bessemer's analysis puts the cost of reaching a validated hit as low as 10,000 dollars, against hundreds of thousands of dollars in traditional workflows. No independent benchmark has been published, and the company did not disclose revenue.
The timing is notable. The round closes days after Isomorphic Labs, Alphabet's drug-design company, raised a 5 billion dollar Series C on November 17, showing that AI drug design is drawing capital at both ends of the openness spectrum. The two playbooks differ sharply: Isomorphic builds its own drug pipeline, while Boltz explicitly refuses to develop drugs, positioning itself as research software that keeps pharmaceutical companies, not a startup pipeline, in control of therapeutic decisions.
For the industry, the interesting experiment is whether pharmaceutical research budgets will sustain usage-based pricing at a scale that covers the compute bill. The open-source-first strategy built the trust and the installed base; the next proof point is expansion inside those six large accounts by mid-2027, when pilots turn into line items that compete with everything else a research team wants to buy.
Comments (0)
Log in to join the discussion
Log InNo comments yet