Bristol Myers Squibb (BMS) has announced the acquisition of an NVIDIA DGX SuperPOD, which it claims is the most powerful and energy-efficient AI supercomputer in life sciences. This announcement comes after Eli Lilly and Roche made their own claims about leading AI supercomputing capabilities within the industry.
BMS's new system is an expansion of its existing infrastructure, which has been in operation since 2024. The new setup will integrate eight racks of Vera Rubin NVL72 systems, featuring 576 GPUs and 36 CPUs per rack, creating a unified computing environment across BMS facilities. This architecture is designed to significantly enhance performance, with BMS citing up to ten times the efficiency per megawatt compared to its previous system.
The competition among these pharmaceutical giants is noteworthy, as BMS, Lilly, and Roche each utilize different architectures. Lilly's system, based on Blackwell GPUs, reportedly offers a peak performance of 4.6 exaflops, while Roche's hybrid-cloud setup includes over 3,500 GPUs but lacks detailed performance metrics. The varying disclosures emphasize the ongoing race for AI supremacy in drug discovery and development.
This escalation in AI capabilities highlights the increasing reliance on advanced computing in the life sciences sector. As companies like BMS aim to leverage AI for drug discovery and clinical applications, the implications for efficiency, cost reduction, and innovation could be transformative for the industry.