Bristol Myers Squibb (BMS) has confirmed plans to build the most powerful AI factory in life sciences by expanding its collaboration with NVIDIA.
The biopharmaceutical company will deploy an NVIDIA DGX SuperPOD powered by DGX Vera Rubin NVL72 systems, creating what it described as the most powerful and energy-efficient single-owned NVIDIA infrastructure in the sector.
The investment gives researchers access to significantly greater computing performance while improving energy efficiency, enabling larger AI workloads across oncology, hematology, cardiovascular disease, immunology and neuroscience.
BMS said the new architecture delivers up to ten times more performance per megawatt than previous systems.
Building on nearly three years of collaboration, the platform will support proprietary AI models that automate target identification, accelerate validation and strengthen the company’s ‘Predict First’ methodology, where AI-generated predictions guide experimental design before laboratory work begins.
BMS said the expanded infrastructure supports its long-term vision of hybrid intelligence, combining AI co-scientists with human researchers to accelerate drug discovery and development. Scientists will use foundation models trained on decades of proprietary research data alongside NVIDIA BioNeMo capabilities to evaluate biological hypotheses, design molecules and improve decision-making throughout development.
Executives from BMS and NVIDIA said the investment would compress discovery timelines, improve research productivity and increase the likelihood of advancing successful medicines into clinical development. The companies believe the enhanced AI infrastructure will help scientists generate higher-confidence insights, reduce manual workloads and speed the delivery of first- and best-in-class therapies for patients worldwide while supporting future advances in precision medicine and biomedical innovation globally.
