Boltz raises $110 million to sell AI drug-design tools to pharmaceutical researchers
Bessemer led the Series A. Boltz reports sharply higher paid usage, but its molecular-design results remain early clues for researchers—not finished medicines.
Loading page…
Bessemer led the Series A. Boltz reports sharply higher paid usage, but its molecular-design results remain early clues for researchers—not finished medicines.
Listen to this story
Boltz is testing whether access to molecular-design models can become a repeat-use research software business: it charges for molecular tokens through Boltz Lab and an API, and reports that paying-customer usage rose 50-fold since its proprietary models launched in June. Its $110 million Series A brings disclosed funding to $138 million and will support computing, data, model development and hiring. The company also reported confirmed hits against six of 10 difficult targets, but these early results are research leads—not drug discoveries or evidence of clinical success.
Bessemer Venture Partners led the round; returning backers include Andreessen Horowitz, while Greycroft, F-Prime Capital, Hummingbird and Benchling joined as new investors.
Boltz says it partners with six of the world’s 20 largest pharmaceutical companies, naming Pfizer, GSK, Takeda and Bristol Myers Squibb.
Boltz reported hits against six of 10 targets after testing 28 to 51 compounds per target; the results do not establish that any candidate is a drug.
Boltz has raised $110 million to help researchers design potential medicines before committing to expensive laboratory experiments. The London company sells AI tools rather than developing its own drug pipeline. Its ambition reaches much further: moving from a therapeutic idea to a molecule ready for human testing through computational design, a goal that remains a substantial scientific challenge.
Bessemer Venture Partners led the Series A, which brings Boltz’s disclosed funding to $138 million following a $28 million seed round unveiled in January. The new financing will support computing infrastructure, data, model development and hiring, according to European Biotechnology Magazine.
Returning investors included Andreessen Horowitz, Amplify Partners, Zetta Venture Partners and Obvious Ventures. Greycroft, F-Prime Capital, Hummingbird and Benchling joined as new backers. Their investment supports a technology-provider model: pharmaceutical and biotech researchers pay to use Boltz’s molecular-design systems in their own research programs.
Instead of charging for individual drug-discovery programs, Boltz sells computational capacity measured in molecular tokens. These units track the molecular information its models process, allowing customers to pay according to usage. Scientists can access the proprietary models through Boltz Lab, a platform for computational experiments, or an application programming interface that connects them to existing research workflows.
Boltz says it has partnerships with six of the world’s 20 largest pharmaceutical companies, naming Pfizer, GSK, Takeda and Bristol Myers Squibb. Its reported growth metric concerns work processed for paying customers, rather than a count of drug candidates or successful treatments.
Founded in 2025 by researchers from the Massachusetts Institute of Technology, Boltz first attracted attention with open-source models. Boltz-1 and Boltz-2 predict molecular structures, interactions and binding affinities—how strongly molecules bind to a target. The company moved into generating molecular designs with two proprietary models launched in June:
Researchers can adapt BoltzMol-1 to practical constraints. A team with a limited supply of compounds can configure it to consider only molecules that can be made from those materials. When procurement is not a constraint, the model can evaluate a larger pool of candidates. That flexibility connects the computational search to what a laboratory can actually make.
In experiments reported in June, Boltz said it found confirmed hits against six of 10 challenging targets while testing 28 to 51 compounds per target. Those figures describe an early research result. As SiliconANGLE explains, small-molecule hits provide clues about where researchers should focus their efforts; they do not amount to drug discoveries on their own.
Boltz has already begun training its next generation of molecular-design algorithms. The company says these are showing substantial improvements over its current models, a company assessment rather than a result from a new external evaluation. The funding will add computing capacity and staff to the business developing those systems.
Loading discussion...
Join the conversation
Explain what adoption proves, and what it leaves unanswered.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.