The pharmaceutical industry's AI moment took two very different shapes today. One is the familiar model — a tech company selling its tools to drugmakers. The other is stranger and more provocative: an AI lab deciding it wants to be the drugmaker itself.
Takeda Puts Real Money Behind AI Discovery
The day's biggest conventional deal comes out of Japan. Takeda, one of Asia's largest drugmakers, has struck a partnership with Insilico Medicine to discover, develop, and commercialize new drug candidates using artificial intelligence. The agreement could be worth up to $600 million.
The structure here is the one the market has come to expect: an established pharma company with deep clinical and commercial muscle pairs with an AI-native biotech that specializes in surfacing promising drug candidates faster than traditional lab work allows. Takeda brings the capital, the pipeline experience, and the path to market. Insilico brings the algorithms that hunt through chemical and biological space for molecules worth pursuing. If the collaboration hits its milestones, it becomes one of the more richly funded validations yet that AI-driven discovery is moving from promise to practice.
Anthropic Wants to Skip the Middleman
The more surprising story belongs to Anthropic, the company behind the Claude AI models. According to reports surfaced by The Verge, ZME Science, and others, Anthropic is moving beyond simply selling its technology to pharmaceutical and research labs — it reportedly wants to develop its own drugs.
The idea, as described, is to put Claude to work as something like a chemist: using the company's AI models to help drive the actual drug development process rather than just supplying tools to the scientists who do. That is a meaningful shift in posture. Most AI firms have positioned themselves as suppliers — arms dealers to the biotech gold rush. Anthropic is signaling it wants a seat at the table as a developer in its own right.
What makes the move more than a business-model curiosity is where it's aimed. Anthropic's program is reportedly targeting neglected diseases — the conditions that big pharma has long treated as too unprofitable to pursue. These are illnesses that affect large populations but sit outside the commercial logic that governs most drug pipelines, because the patients who need treatments can't sustain the returns investors expect. By pointing its effort at exactly the diseases the industry avoids, Anthropic is framing this less as a land grab and more as an attempt to use cheaper, faster AI-driven research to make previously uneconomical work viable.
Two Bets, One Question
Put side by side, the two stories map the fault line running through AI in medicine right now. Takeda and Insilico represent the collaboration model — established players and AI specialists sharing risk and reward through a big, milestone-based partnership. Anthropic represents the vertical-integration bet: if your models are good enough to accelerate discovery, why hand that value to someone else instead of capturing it yourself?
Both bets rest on the same unproven premise — that AI can genuinely compress the brutal timelines and costs of drug development. Takeda is wagering up to $600 million that it can. Anthropic is wagering its own credibility that Claude can do enough of the scientific heavy lifting to justify becoming a drugmaker, and that it can do so for diseases the market has written off.
Neither approach has delivered an approved medicine yet, and the distance between a promising AI-generated candidate and a drug on a pharmacy shelf remains long and littered with failure. But today made the stakes clearer. The tool-sellers are cashing in, and at least one of them has decided the tools might be worth more if it uses them itself. Watch which model produces a real therapy first — that's the answer that will reshape how this industry spends its next decade.