Anthropic is reported to have acquired Coefficient Bio, a New York biotech AI startup, in a stock deal valued at about $400 million, according to The Information. TechCrunch separately reported that sources close to the transaction confirmed to it that the deal had closed, although they declined to comment on the amount.
Who Coefficient Bio’s founders are and what they built
Nathan C. Frey, who identifies himself publicly as Coefficient Bio’s CTO and co-founder, says on his personal site that he was previously a group leader and principal scientist at Prescient Design at Genentech, where he led a multidisciplinary team working on biological foundation models and AI and machine learning approaches to biomolecule design.
Frey also says he sat on Roche and Genentech leadership teams for foundation models and large-molecule drug discovery, established the group’s collaboration with NVIDIA and has published more than 20 scientific papers across journals and conferences including Science Advances, Nature Machine Intelligence, NeurIPS, ICML and ICLR.
Public research and code repositories also tie the Coefficient founders to biology-native model work in protein design and sequence modeling. On OpenReview, Samuel Don Stanton and Nathan C. Frey are listed as coauthors of Protein Design with Guided Discrete Diffusion, a NeurIPS 2023 spotlight paper on protein design.
On GitHub, the Prescient Design repository for the biological sequence modeling library LBSTER (or Lobster) says the project was led by Nathan Frey across Prescient Design and Genentech, and that it was built with Beignet and integrated with Cortex, which the repository describes as a modular framework for multitask modeling, guided generation and multimodal models.
Samuel Stanton’s public GitHub profile identifies him as technical staff at Coefficient Bio and formerly an ML scientist at Prescient Design.
The life sciences program Anthropic had already built
The background also fits a life sciences agenda Anthropic had already laid out before the reported deal surfaced. When it launched Claude for Life Sciences in October 2025, Anthropic said its goal was to make Claude capable of supporting “the entire process, from early discovery through to translation and commercialization.”
The company said the product was aimed at researchers, clinical coordinators and regulatory affairs managers. In January, Anthropic said it was expanding those capabilities further, adding support tied to scientific platforms and to work spanning clinical trial management and regulatory operations.
Placed against that record, the reported Coefficient Bio transaction reads less like a first step into life sciences than an attempt to add a more specialized research team to a program Anthropic had already formalized. Anthropic’s October and January materials described product integrations, workflow support and domain-specific use cases.
Frey’s public biography and the founders’ published work point to experience in biological foundation models, generative modeling and protein design. The reported deal would put those strands inside the same company.
How the deal fits Anthropic’s recent acquisition pattern
The move would also fit a broader acquisition pattern Anthropic has described in its own announcements. In December, the company said it had acquired Bun to accelerate Claude Code.
In February, it said it had acquired Vercept to improve Claude’s computer-use capabilities. Both announcements framed the deals as ways to bring specialized technical teams and infrastructure closer to Anthropic’s core products.
If the Coefficient Bio deal is read alongside those company statements, it would mark the same kind of build-buy decision in a different domain: life sciences.
The broader market context the deal lands into
The wider market context also gives the deal more weight than a stand-alone acqui-hire story. In January, NVIDIA and Eli Lilly said they would jointly invest up to $1 billion over five years in a co-innovation lab for drug discovery, bringing together Lilly’s biology and medicine teams with NVIDIA’s model builders and infrastructure.
Separately, the FDA said in May 2025 that its first AI-assisted scientific review pilot had shown promise, with one FDA reviewer saying it enabled scientific review tasks to be completed in minutes rather than the three days they previously took, and in June launched Elsa as an agency-wide AI tool, saying it planned to integrate more AI into agency processes over time.