Anthropic is doing more than selling tools to drugmakers; it is setting up to do the drug discovery itself. The AI developer is advancing a biology laboratory of its own, according to a Reuters report on September 18, adding wet-lab experimentation to a drug discovery program that has until now existed mostly in software.
The ambition has been telegraphed for months. When Anthropic released Claude Science on June 30, it said it intended to pursue drug discovery directly, and it chose a distinctive set of targets: tuberculosis, malaria, and other diseases that large pharmaceutical companies largely avoid because the commercial returns are thin. The choice signals a program aimed at neglected diseases rather than blockbuster markets.
The pieces were assembled quietly. In April, Anthropic bought Coefficient Bio, an AI biotech startup, for about $400 million in an all-stock deal, acquiring a team that had been building machine-learning tools for biology. The company also hired Eric Kauderer-Abrams, a former Verily executive, to lead its life sciences effort. On September 17, it launched a life sciences validation program, another step in building the infrastructure to test its own discoveries.
The decision to run a lab is a departure from the usual playbook. AI companies have preferred to partner with pharmaceutical firms, selling models and software rather than taking on the cost and complexity of experimental biology. A wet lab, with its benches, reagents, and regulatory obligations, is a different kind of business, and Anthropic is choosing to operate one while continuing to sell its tools to the same industry.
The logic is that the two activities reinforce each other. Running experiments generates data that improves the models, and the models in turn suggest better experiments. A company that does both closes a loop that a software-only vendor cannot, and the lab gives Anthropic a way to validate its own technology that partnership alone would not provide.
The drug discovery market is one of the largest opportunities in applied AI. The process of finding a drug is slow and expensive, and the hope is that AI can compress it. Anthropic’s entry positions it against a field of well-funded competitors, from established pharma companies building their own AI groups to startups pursuing the same goal.
The neglected-disease focus is unusual and may limit the commercial upside, at least initially. Tuberculosis and malaria are not lucrative targets, and the funding for them comes largely from governments and philanthropies. But the choice also reduces competition and aligns the program with a public-interest narrative that has been part of Anthropic’s identity since its founding.
Anthropic has not disclosed how large the lab will be, where it is located, or how much it will cost to run. The Reuters report described the effort as advancing, and the company has confirmed its drug discovery ambitions without detailing the timeline. The hiring and the acquisition suggest the program is already staffed and funded.
The wet lab is the piece that separates Anthropic’s program from a software project. Discovering a drug on a computer produces a hypothesis; turning it into a medicine requires testing it in cells, animals, and eventually people, a process that cannot be done in silico. A laboratory gives Anthropic a way to run those experiments itself, closing the loop between its models and the physical world they are trying to influence.
The field of AI-driven drug discovery has drawn large amounts of capital, and its results so far have been mixed. Several companies have put AI-designed molecules into clinical trials, but the long, uncertain path from discovery to an approved drug has tempered early enthusiasm. Anthropic’s program is entering a field where the technology is promising and the record is still being written.
The neglected-disease focus shapes the program’s economics. Tuberculosis and malaria are not lucrative targets, and the funding for them comes mostly from governments and philanthropies rather than commercial markets. That choice limits the financial upside but also reduces competition and aligns the work with a public-interest mission. It is a deliberate positioning, and it says something about what Anthropic wants the program to be.
The company has described the tools and the drug program as complementary rather than competing, and the validation program launched on September 17 appears designed to test that premise. Selling models to other researchers while running its own experiments lets Anthropic learn from both sides of the transaction, at the cost of competing, in some measure, with the customers it also wants to serve.
The broader question is whether an AI company can operate a pharmaceutical operation without being pulled off course. The two businesses have different rhythms, different regulators, and different definitions of success, and the industry has watched other technology companies struggle to enter healthcare. Anthropic’s decision to do both at once, selling tools and making drugs, is a bet that it can manage both, and that the loop between them is worth the cost.
