The part that sticks with me isn’t the “AI will transform medicine” flourish. It’s that Novo is doing this while it is also in the middle of a rebrand, a strategy reset, and a very public race to look more modern than Eli Lilly. That makes the Anthropic partnership feel less like a pure science story and more like a signal to investors, competitors, and maybe its own employees: we’re going to be an AI company now, or at least talk like one.
I’m not dismissing the deal. If you’re running a giant drugmaker, there’s obvious value in using frontier models to help scientists sift through biological messes, write tooling, and speed up repetitive workflows. That part sounds sensible. What I’m less convinced by is the usual grand language around “compressing a century’s worth” of breakthroughs into a decade. That kind of line is cheap. It flatters the vendor, flatters the customer, and tells you almost nothing about where the real bottlenecks are. Drug discovery is not short on text generation. It is short on validated biology, clean data, experimental throughput, and the ability to tell a promising hypothesis from a dead end.
The more interesting detail is that Novo is already using OpenAI, and now it’s adding Anthropic too. That tells me these companies do not want to bet on one model provider for something this sensitive. Fair enough. In practice, pharma probably wants a portfolio of models and a lot of internal control around them. If I were building in this space, that’s the part I’d pay attention to: not which chatbot got the press release, but how the workflows are actually partitioned, what data stays internal, and whether the models are being used for real scientific reasoning or just for faster scaffolding around human work.
I also noticed the tension in the article itself. Anthropic is talking about safe, trusted frontier models, while the same company has recently had to explain misuse around surveillance and weapons. Novo says there will be robust governance and human oversight. Good. But those phrases are doing a lot of heavy lifting. In regulated industries, “human oversight” can mean anything from meaningful expert review to a checkbox on a procurement slide. I’d want to know who can audit outputs, how errors are caught, and whether any of this touches decisions that matter for patient risk.
Still, this is a believable place for Claude to show up. Not as the magical molecule machine, but as the layer that helps scientists and computational teams move faster on the ugly, tedious parts of R&D. If it ends up saving serious time there, that will be useful even if the marketing is inflated. The real test is whether Novo can show something beyond vibes: fewer dead-end experiments, better prioritization, or faster internal tooling that scientists actually keep using.
Reference: Novo teams up with Anthropic to speed up drug discovery
