Medical AI Startup Boltz Raises $110M Series A For Drug Design

Boltz PBC, a London-based startup developing artificial intelligence models for pharmaceutical researchers, has raised $110 million in Series A funding led by Bessemer Venture Partners.
Following a $28 million seed financing disclosed in January, the Series A lifts Boltz's total announced funding to $138 million, Superpower Daily reported. Returning backers Andreessen Horowitz, Amplify Partners, Zetta Venture Partners, and Obvious Ventures joined the round alongside new investors Greycroft, F-Prime Capital, Hummingbird, and Benchling, according to SiliconANGLE and Superpower Daily. Established in 2025 by Massachusetts Institute of Technology researchers, Boltz markets AI software to biotech and pharma firms rather than building an internal drug pipeline, Superpower Daily noted. Its broader goal is to advance computational design from a therapeutic concept to a molecule ready for human testing, which Superpower Daily described as a major scientific hurdle. The hosted Boltz API is priced from as little as $0.025 per prediction, according to Labcritics. On January 8, 2026, MIT Jameel Clinic project Boltz incorporated as a public benefit corporation, launched Boltz Lab, and announced a partnership with Pfizer, according to Community Jameel.
Molecular Models And Early Research Hits
Boltz first gained attention with open-source models Boltz-1 and Boltz-2, which predict molecular structures, interactions, and binding affinities, before launching proprietary models BoltzMol-1 and BoltzProt-1 in June, Superpower Daily reported. Traditional drug discovery screens hundreds of thousands of candidate molecules for binding affinity with pathogens, a workflow that can take months and cost millions of dollars, SiliconANGLE explained. BoltzMol-1 narrows that search to a few dozen promising candidates, and researchers can constrain it to molecules synthesizable from a lab's available compounds or let it scan a wider pool when procurement is unrestricted. In June experiments across 10 "challenging and therapeutically-relevant targets," Boltz reported confirmed small-molecule hits for six targets after testing 28 to 51 compounds per target, and Superpower Daily noted that these hits are early research leads rather than finished drugs. Its companion model, BoltzProt-1, generates tens of thousands of protein designs from hit data and evaluates each candidate's suitability for drug development. Whereas competing models rely on indirect rubrics, BoltzProt-1 uses direct metrics that Boltz says achieved "higher success" against rivals in internal testing, SiliconANGLE reported.
Usage Growth And Pharma Partnerships
Customers access the proprietary models through the Boltz Lab platform or an application programming interface billed by molecular tokens, with token consumption rising 50-fold since June, SiliconANGLE and Superpower Daily reported. Boltz counts six of the 20 largest pharmaceutical companies as partners, including Pfizer, GSK, Takeda, and Bristol Myers Squibb, Superpower Daily noted. The company will use the new capital to expand computing infrastructure, data, model development, and hiring as it competes with biology-focused models from newly public Anthropic PBC and OpenAI Group, while training next-generation algorithms that Boltz says show "substantial improvements." Its tools are also deployed at fragrance and ingredient company DSM Firmenich, according to Bessemer Venture Partners.
