Everything to Know about BMS & NVidia’s AI Supercomputer

Bristol Myers Squibb (BMS) has announced plans on July 20, 2026 to expand its computing infrastructure with NVIDIA DGX Vera Rubin NVL72 systems, a move the company says will give it the most advanced and energy-efficient artificial intelligence (AI) infrastructure of any single owner in life sciences.¹ The deployment extends a collaboration that began nearly 3 years ago, when BMS first adopted NVIDIA DGX SuperPOD systems to support its research and development work.

The Vera Rubin architecture is designed to deliver up to 10 times greater performance per megawatt than its predecessor.¹ That efficiency gain matters as much as the raw processing boost, because it allows BMS to pursue more complex AI models without a proportional rise in energy demand. The expanded infrastructure will support research programs across oncology, hematology, cardiovascular disease, immunology, and neuroscience.

Why Does Computing Capacity Matter to Drug Developers?

Pharmaceutical development increasingly depends on the ability to train and run proprietary models on large volumes of internal experimental and clinical data.¹ The scale of that computing capacity shapes how quickly companies can test hypotheses, design molecules, and interpret results, which in turn affects how efficiently a program moves from discovery toward clinical development. As more of the industry builds this kind of dedicated capacity, computational infrastructure is becoming a factor in development timelines alongside laboratory and clinical execution.

BMS points to 2 examples of this shift already underway.¹ AI agents used for target identification and validation are reducing manual work for scientists. Separately, the company's Predict First approach uses model-generated predictions to inform experimental design before laboratory work begins, a method it says now informs every small molecule program and most large molecule programs it runs.

What Role Is AI Expected to Play Alongside Researchers?

BMS describes its broader goal as hybrid intelligence, an approach in which computational systems handle data-intensive execution while human researchers retain responsibility for direction, interpretation, and judgment calls that require deeper expertise.¹ The Vera Rubin cluster is intended to serve as the computational foundation for that model, supporting foundation models trained on the company's proprietary research data and, where applicable, incorporating domain-specific tools from NVIDIA's biological AI platform, BioNeMo.

“Drug discovery is a sequence of decisions made under uncertainty, and better decisions come from better evidence, faster,” said Robert Plenge, executive vice president and chief research officer, Bristol Myers Squibb, in a press release.¹ “This infrastructure lets us learn from every experiment and every clinical readout to sharpen the next hypothesis, allowing BMS scientists to spend less time on manual work and more time on the questions that require human judgment. The goal isn't speed for its own sake; it's raising the probability that each program we advance is the right one.”

The significance lies less in the hardware itself and more in what it signals about where competitive advantage is shifting.¹ As computational tools become embedded earlier in discovery and design, organizations that can evaluate more hypotheses, faster and with higher confidence, may be positioned to bring a narrower, better-vetted set of candidates into costly development and manufacturing pipelines. That could influence how resources are allocated downstream, from process development through scale-up, as fewer but more thoroughly validated programs move forward.

How Big Could This Infrastructure Shift Become?

BMS is not moving in isolation.² Analysts at Global Reports Store value the global AI drug discovery infrastructure market at $3.10 billion in 2025, projecting growth to $19.98 billion by 2032. That trajectory reflects a broader shift: pharmaceutical research is being rebuilt around compute, data, and generative molecular design rather than isolated software tools. Vendors and internal groups are increasingly judged on whether their data architecture, computing environment, and lab integration can make artificial intelligence useful at production scale, not just in early discovery.

References

1. Bristol Myers Squibb to build the most powerful AI factory in life sciences with NVIDIA. Press release. Bristol Myers Squibb. July 20, 2026. https://news.bms.com/news/corporate-financial/2026/Bristol-Myers-Squibb-to-Build-the-Most-Powerful-AI-Factory-in-Life-Sciences-with-NVIDIA/default.aspx
2. Global Reports Store. AI Drug Discovery Infrastructure Market to Add US$16.88 Billion by 2032 as Pharma, Cloud and Bio-AI Platforms Rebuild R&D Around Compute, Data and Generative Molecular Design. openPR. May 13, 2026. https://www.openpr.com/news/4512781/ai-drug-discovery-infrastructure-market-to-add-us-16-88-billion