Bristol Myers Squibb deploys second DGX SuperPOD on Vera Rubin; doubles AI compute for unified drug discovery platform

Bristol Myers Squibb announced July 21 that it is deploying its second NVIDIA DGX SuperPOD, built on eight DGX Vera Rubin NVL72 systems. The new cluster delivers up to 10x performance per megawatt compared to the infrastructure it replaces, providing unified AI infrastructure for the pharmaceutical giant's research operations. BMS is opening compute access to its entire researcher base—no queuing, no artificial limits—enabling workflows across the full drug discovery pipeline.

BMS's first SuperPOD (deployed ~3 years ago) has already produced measurable wins: AI-enabled target identification saves scientists weeks of manual work; the company expanded its CELMoD compound library (molecules engineered to degrade cancer-causing proteins); and a "Predict First" methodology uses AI predictions to gate experimental synthesis, accelerating lead optimization. Teams integrate datasets from facilities across the US and globally into a single learning loop.

The new Vera Rubin-powered system unifies two separate clusters into a single data plane accessible from every BMS site globally. NVIDIA Mission Control manages the infrastructure, letting researchers initiate complex predictions in plain English rather than navigating computational logistics. The company is building its own foundational models, demanding significant GPU capacity.

For biotech planners: this is a production snapshot of the inference + training + agentic-workflow stack in scale. BMS views compute as non-fungible infrastructure that compounds learning across programs—a shift from treating projects as discrete experiments. Watch whether other pharma majors follow this model of unified, researcher-facing compute rather than allocating resources per program.