For most of the past decade, artificial intelligence in pharmaceuticals looked like a software story. Labs like OpenAI and Anthropic would build the tools; drugmakers would pay to use them. The acquisitions of the past several months suggest that model is being abandoned before it fully arrived.
OpenAI has acquired Orchestra Bio, a platform that helps biopharma teams coordinate the complex process of drug development, and is bringing Orchestra’s team and technology together with Rosalind, its AI models and tools for life sciences. Orchestra builds AI agents intended to help pharmaceutical companies manage drug-development operations. The financial terms were not disclosed. Orchestra had raised around $20 million as of 2025 from backers including Andreessen Horowitz, Atria, and 8VC.
The acquisition is not a standalone event. Anthropic acquired AI biotech startup Coefficient Bio for roughly $400 million in April. The deal brought a team of fewer than ten people into Anthropic’s healthcare and life sciences division, with cofounders whose backgrounds include Genentech’s Prescient Design computational drug discovery unit. According to The Information, Coefficient Bio had built a platform that enabled AI to carry out biotech tasks including drafting drug R&D plans, managing clinical regulatory strategies, and discovering new drug candidates. Before Orchestra, Axios reported that OpenAI acquired Torch, a health tech startup that unifies lab results, medications and recordings from doctor visits.
Three acquisitions in ten months. Combined disclosed and estimated consideration north of half a billion dollars. The pattern is unmistakable: these companies are not selling software to pharma. They are becoming pharma-adjacent operators themselves, owning the workflow layer that sits between discovery and the clinic.
That distinction matters enormously for long-term investors. Consider what it means for the incumbents. IQVIA, the clinical research giant trading as IQV, announced in February that it signed an agreement to acquire certain discovery service assets from Charles River Laboratories, including five sites specializing in in vitro drug discovery services and a small molecule AI platform. IQVIA paid approximately $145 million for those assets. That is a company defending its territory. Veeva Systems, which dominates pharma software with its cloud CRM and data tools, has spent years embedding itself as the connective tissue of drug commercialization. Both businesses built durable switching costs precisely because replacing their workflows is painful and slow.
The question now is whether OpenAI and Anthropic are building something stickier, or something that merely competes at the margin. OpenAI has said that developing a new medicine typically takes 10 to 15 years, a timeline it hopes AI can shorten. OpenAI has disclosed partnerships with Novo Nordisk around applying its AI capabilities across the drug lifecycle. Rosalind, the life sciences suite absorbing Orchestra’s capabilities, brings AI models and tools to life sciences researchers, with GPT-Rosalind aimed at research across biology and drug discovery, and Rosalind Workbench positioned as an orchestrated workspace that connects those capabilities with scientific tools and data to help researchers analyze results and plan experiments.
That is not a chatbot bolted onto a pharma website. It is an attempt to own the operating layer of drug development, from hypothesis to trial coordination. If it works, the value does not accrue to the pharma company paying a software subscription. It accrues to the AI lab sitting between the science and the decision.
The risk is real. Drug development is regulated, slow, and littered with the wreckage of companies that believed technology could compress its timelines on a predictable schedule. Execution risk for organizations with no history of navigating FDA submissions or managing CRO relationships is not trivial. And the incumbents, IQVIA and Veeva in particular, have decades of data, relationships, and workflow integrations that no acquisition of a 50-person startup can replicate overnight.
The question of who captures value in drug development is not limited to AI labs and incumbents. Chinese biopharma out-licensing has restructured where innovation originates and who keeps the economics when a molecule reaches Western markets.
But the strategic direction is clear enough to track now, before the revenue arrives. When two of the best-capitalized technology companies on earth start buying the drug-development process rather than licensing tools to it, the question is not whether the industry changes. It is who captures the economics when it does.
