Bristol Myers Squibb

Bristol Myers Squibb is a global biopharmaceutical company headquartered in Princeton, New Jersey (NYSE: BMY). It is best known for the cancer immunotherapy Opdivo, the blood thinner Eliquis (developed with Pfizer), the multiple myeloma drug Revlimid, the cell therapy Breyanzi and the schizophrenia medicine Cobenfy. It appears here not as an AI vendor but as a large pharmaceutical AI adopter. It sells no AI product; it uses AI inside research, clinical development, manufacturing and its commercial operations.

Its defining pattern is the full-stack adopter: it owns the compute, standardizes the model layer, and buys specialist tools on top of both. The compute came first. The company has run its own NVIDIA DGX SuperPOD for several years, and in July 2026 it said it would expand to a new SuperPOD built on NVIDIA's Vera Rubin systems, which it describes as the most powerful single-owned NVIDIA infrastructure in life sciences. The model layer followed in May 2026, when it signed a strategic agreement with Anthropic to make Claude Enterprise the shared intelligence platform for more than 30,000 employees. Engineering and data science teams use Claude Code to build software, and the company names drafting clinical study reports and patient safety narratives from trial data among the target uses. The same announcement calls this part of a deliberate multi-vendor strategy: for several years, staff have reached frontier models from several labs through a proprietary internal platform.

The specialist layer is a run of 2026 deals. In January it teamed with Microsoft to deploy radiology algorithms cleared by the US Food and Drug Administration through Microsoft's imaging network, flagging lung nodules on X-rays and CT scans and tracking patients who would otherwise be lost to follow-up. In February it agreed to run the Cost Optimizer module of Study Designer, from the AstraZeneca-owned clinical technology company Evinova, across its global trial portfolio. In August it signed on to deploy Bunsen, Schrödinger's agentic AI co-scientist, across its research organization, extending a long relationship with that company, and it began a collaboration with Chai Discovery to design antibody candidates with Chai's folding and design models. An older partnership continues alongside them: since 2020 insitro has used machine learning to find targets for amyotrophic lateral sclerosis (ALS), and in March 2026 Bristol Myers Squibb nominated two more. No financial terms were disclosed for the Schrödinger or Chai deals. Worth being precise: Schrödinger's core platform is built on physics-based simulation, with machine learning alongside it, and Bunsen is the agentic layer that plans and interprets that work.

The honest framing: Bristol Myers Squibb's value rests on its marketed medicines and its pipeline, and AI is a lever on the cost and speed of finding and developing the next ones. Its chief digital officer says the bet is beginning to pay off, but none of these announcements publish a measured result, so the scale of the stack is the fact and the return is still the intention. Among peers in the catalog, Gilead procures specialists without owning the compute, Eli Lilly productized its own models through TuneLab, Moderna built hundreds of models in-house, Merck & Co. runs one enterprise platform with Google Cloud, and Novo Nordisk also signed with Anthropic but as one of several vendors alongside OpenAI and Amazon Web Services. Bristol Myers Squibb is multi-vendor too; what sets it apart is that it owns the supercomputer underneath.