Bristol Myers Squibb will become the first life-sciences company to purchase an Nvidia DGX SuperPOD built on the new Vera Rubin architecture, a bet the drugmaker frames as a decisive expansion of AI's role in discovering medicines. It calls the machine the most powerful single-owned Nvidia infrastructure in life sciences.
A step-change in efficiency
The DGX Vera Rubin NVL72 system offers up to 10x greater performance per megawatt than its predecessor, a SuperPOD that BMS has run since March 2024. Chief Digital and Technology Officer Greg Meyers put it plainly: the company gets "10 times more compute capacity per watt spent," adding that "BMS has made a deliberate bet on AI, and we are beginning to see it pay off."
Where the compute goes
BMS will run proprietary AI models across oncology, hematology, cardiovascular, immunology and neuroscience programs. Chief Research Officer Robert Plenge framed the payoff in throughput: on evaluating drug candidates, "maybe before we could do 10 and now we can do dozens." The compute lets the company test far more molecular hypotheses in silico before committing to lab work.
Measurable time savings
The company says AI has already cut its time-to-clinical-testing by 20-30%, and it is targeting 50%. More striking is its claim that an experimental sickle cell disease treatment would not have been discovered without AI-enabled research — a concrete example of a candidate that existing methods missed. Financial terms of the Nvidia purchase were not disclosed.
Owning the iron
By buying rather than renting cloud capacity, BMS keeps its models and data on infrastructure it controls — a recurring pattern among enterprises handling sensitive, high-value IP. For pharma, where a single approved drug can be worth billions, owning the compute that shortens the discovery cycle is increasingly treated as core infrastructure rather than an experiment.