Novo Nordisk enlists Anthropic's Claude Science for drug discovery, no terms disclosed

NVO and Anthropic said on September 16 that they will collaborate to apply AI to drug discovery, with Novo initially testing Anthropic's Claude Science application and frontier models against research problems identified by its own scientists and computational teams. The scope also extends beyond the laboratory to agentic software engineering across the company, and the announcement says the work will operate under what the companies describe as robust data governance and human oversight. What the announcement does not contain is a price: no dollar value, no contract term and no headcount were disclosed, and both sides describe this as an initial testing phase rather than a committed multi-year programme.

Claude Science is an Anthropic application rather than a new model. By Anthropic's own description it layers a coordinating agent and a library of curated scientific skills over existing Claude models, with connectors into scientific databases and life-sciences tooling spanning genomics, proteomics, structural biology and cheminformatics. That distinction matters for how quickly the partnership could show results: Novo is adopting a workflow layer already built for scientific tasks, not waiting on a bespoke model to be trained on its data.

This is the third named AI partnership Novo has announced in 2026, following OpenAI in April and AWS in August, the latter covering more than 25,000 employees. Novo already runs NovoScribe, an internal Claude-based platform for clinical and regulatory documentation that predates this agreement, so Claude is not new inside the company; what is new is pointing it at discovery rather than paperwork. Chief Executive Mike Doustdar framed the deal as part of an ambition to become what he called the world's most AI-driven healthcare company, while Anthropic's Dario Amodei repeated his argument that AI could compress a century of biological and medical breakthroughs into a decade.

The market treated it as a non-event. Novo shares were essentially flat on the announcement and changed hands at $41.71, down 1.9% on the session, which is a reasonable read given that an untermed research collaboration produces no near-term revenue and Novo's pressing problem remains competition in obesity rather than discovery throughput. SentiSense data does show the story landing in the conversation even where it did not land in the tape: sentiment on the ticker spiked to +0.70 against a 30-day average of +0.17, a reading 2.3 standard deviations above its recent baseline. The checkable milestones from here are narrow and worth naming: whether the initial testing phase converts into a disclosed commercial agreement, and whether any named discovery programme is attributed to the collaboration.