Bristol Myers Squibb Partners With NVIDIA to Build Pharma's Largest Supercomputer

American biopharmaceutical company Bristol Myers Squibb (BMS) has expanded its technology partnership with chipmaker NVIDIA Corporation (NVIDIA) to construct a massive supercomputing cluster. The pharmaceutical giant plans to deploy a second NVIDIA DGX SuperPOD built on eight DGX Vera Rubin NVL72 rack-scale systems to accelerate early drug discovery operations.

This technical deployment gives the biopharmaceutical firm what it claims is the most powerful and energy-efficient single-owned artificial intelligence (AI) infrastructure in life sciences. The upcoming facility will incorporate 576 Rubin graphics processing units (GPUs) alongside 288 Vera central processing units (CPUs) that function as a single unified computing system.

Engineers designed the next-generation architecture to deliver up to ten times greater performance per megawatt than the infrastructure it replaces, which helps manage rising electricity costs as computational demand grows. Executive leadership confirmed that the installation will integrate with an existing supercomputer cluster deployed three years ago, which reached full capacity.

Bristol Myers Squibb Chief Digital and Technology Officer Greg Meyers stated that expanding hardware capabilities provides researchers with the scale required to maintain an industry lead as computational algorithms expand. He emphasized that the financial investment reflects a long-term strategy that is already yielding measurable benefits across active clinical drug pipelines.

The unified computing environment will support internal research teams working across five core therapeutic divisions, which include oncology, hematology, cardiovascular disease, immunology, and neuroscience. Scientists plan to utilize the high-speed hardware to train proprietary foundation models on decades of internal scientific data, while drawing upon specialized tools through the NVIDIA BioNeMo software platform.

Under its established Predict First operational methodology, the pharmaceutical group uses automated predictions to guide experimental design before physical work begins in laboratory settings. Company officials confirmed that computer forecasts currently inform every small molecule drug program, while also guiding a substantial majority of large molecule development initiatives across active units.

Bristol Myers Squibb Chief Research Officer Robert Plenge explained that the additional computing power allows scientific teams to evaluate dozens of potential medicine candidates simultaneously rather than a limited few. He noted that the primary goal is not speed alone, but increasing the overall probability that selected molecules perform successfully during clinical trials.

The expanded system will also run automated agentic workflows, which allow virtual assistant software to process research data across traditional departmental boundaries without human delay. Research Business Insights and Technology Vice President Erin Davis indicated that the platform will be accessible from every corporate facility globally, allowing scientists to launch complex analysis tasks using plain commands.

Competing drug manufacturers like Eli Lilly and Company and Roche Holding AG have similarly invested in dedicated hardware clusters to speed up therapeutic research. Industry analysts note that owning dedicated supercomputers prevents reliance on third-party cloud capacity, while offering tighter protection over proprietary biological data during complex molecular modeling operations.

The project represents a growing trend where major pharmaceutical firms build dedicated technology factories to process massive internal datasets. By moving away from shared cloud arrangements, drugmakers can run continuous computational predictions without encountering system throttles or bandwidth constraints during peak development cycles.

This hardware installation complements other recent technology agreements signed by the pharmaceutical giant, including an enterprise-wide rollout of generative software models to more than 30,000 employees. Company officials anticipate that the expanded supercomputing infrastructure will become fully operational across its global research network by January 2027.