I said Anthropic would carry no drug asset. I was wrong

Anthropic filed for its IPO on Monday. On Tuesday it disclosed a drug program it bought from a one-person startup. In August I wrote it would carry no asset on its balance sheet. I was wrong.

Tuesday's update names six partners (BigHat, Bristol Myers Squibb, Divergence Labs, Enveda, Generate, The Column Group) and one acquisition: Mithril Medicine, a company of one. You probably won't have heard of it unless you were very deep into the X techbio chatter.

Its founder and CEO, Cody Tranbarger ran contracts, logistics and experiments across 11 external labs with Claude and took a candidate for a rare muscular dystrophy from idea to mouse data in under 18 months. He now works at Anthropic. A first human trial could start in late 2028. A one-employee biotech running a program across 11 labs was unthinkable before AI. The Column Group is building on the same idea: Pleiades, a new company set up to run far more programs on Claude than a small team could.

There is also a partnership targeting manufacturing and regulatory, with Enveda. Claude read 1,061 pages of batch documentation, found all 11 issues the reviewers had flagged plus seven they missed, and cut review time by 89%, from over 25 hours to under three. A regulatory document that takes about three months by hand came back as a first draft in four hours.

Now, to what I make of all this. In August I said Anthropic can work on thousands of programs without carrying a single asset on its balance sheet. That line is now clearly wrong. It owns programs in two rare muscular dystrophies, and the S-1 lists its own drug candidates as a risk that may never produce revenue.

Back then I also predicted that pharma keeps carrying the clinical risk from Phase II onward. Neither the filing nor the update touches it. In September Anthropic's life sciences head told Reuters it would not run clinical trials (for now).

In September I predicted that Anthropic will not start buying pharma companies or CROs, and the bar for owning rather than partnering rises once it has public shareholders. This still holds, at least to what I can find in the S1. The one acquisition disclosed in IPO week has a headcount of one. The six named companies are partners.

So for now, to big pharma, Anthropic remains mostly a partner, and is not yet competing directly. I still think it will stay that way, but I also presume any head of R&D in pharma is going to think twice before opening its books too deep to Claude going forward.