Princeton-based Biopharmaceutical leader Bristol Myers Squibb (BMS) announced an expansion of its computing infrastructure through an agreement with NVIDIA to deploy an NVIDIA DGX SuperPOD powered by DGX Vera Rubin NVL72 systems.
The deployment establishes a high-performance computational environment for the company’s research operations, using NVIDIA’s latest rack-scale architecture.
The characterization of the system as “the most powerful single-owned infrastructure in the sector” originates directly from corporate and partner statements issued by BMS and NVIDIA. These claims are based on specific architectural benchmarks, most notably the Vera Rubin system’s capability to deliver up to ten times greater performance per megawatt compared to previous-generation infrastructure.
Rather than relying on raw superlatives, the technical significance lies in the scale of the deployment:
- Energy-Efficient Scaling: The architecture allows computational teams to process significantly larger and more sophisticated AI workloads without incurring a proportional surge in power consumption.
- Advanced Hardware Integration: By deploying eight rack-scale systems featuring NVIDIA Vera CPUs and Rubin GPUs, the infrastructure provides a unified platform designed to support heavy model training and agentic workflows.
The infrastructure serves as the computational backbone for BMS’s ongoing digital initiatives across oncology, hematology, cardiovascular health, immunology, and neuroscience. By integrating platforms like NVIDIA’s BioNeMo toolkits, the system supports automated target validation and the company’s “Predict First” methodology, which uses predictive modeling to guide experimental design prior to laboratory execution.
