BMS and NVIDIA Build Pharma’s AI Supercomputer

BMS and NVIDIA are scaling a new AI factory for drug discovery, combining NVIDIA Vera Rubin computing with BMS’s proprietary scientific data.

It’s not just the size of the supercomputer, but also the ability to turn high-performance computing, foundation models, and AI agents into a shared scientific infrastructure across the drug discovery pipeline.

Image

Why BMS Is Building a Pharma AI Factory

Bristol Myers Squibb is expanding its collaboration with NVIDIA with a new DGX SuperPOD built around eight DGX Vera Rubin NVL72 systems.

This will undoubtedly be the largest and most powerful supercomputer in pharma.

BMS says the infrastructure will provide its researchers with its most powerful and energy-efficient single-owned NVIDIA computing environment in life sciences.

The system is designed for workloads that conventional enterprise computing cannot handle at the same scale. These include training foundation models, running large predictions, exploring chemical space, and supporting agentic workflows.

More broadly in the industry, we’ve seen AI moving from isolated experiments into core pharmaceutical R&D.

AI applications in drug discovery span target identification, drug discovery, preclinical research, clinical development, and post-market surveillance. 

BMS is therefore building infrastructure around this idea with a clear focus on what happens when AI becomes available as a shared scientific capability rather than a specialist tool.

Read Reimagine AI-driven Drug Discovery with Pharma Superintelligence

Reimagine AI-driven Drug Discovery with Pharma Superintelligence

Explore how AI-driven drug discovery is evolving through Pharmaceutical Superintelligence and what your organisation can do to keep up.

From Pharma Supercomputer to Scientific Infrastructure

The new BMS NVIDIA supercomputer is not intended to operate as a standalone research machine.

BMS plans to combine the new Vera Rubin environment with its existing NVIDIA infrastructure into a more unified computing environment.

This is designed to give researchers across locations access to models, data, and computational resources without the same organisational and technical barriers that can fragment pharmaceutical R&D.

The underlying architecture is very important, as drug discovery generates multiple forms of information. Experimental results, molecular structures, clinical observations, imaging, and other proprietary datasets can become more valuable when they are connected.

BMS is also developing AI models trained on its proprietary scientific knowledge, while using NVIDIA BioNeMo capabilities for biological AI applications.

Nvidia BioNeMo supports model development across areas such as protein structure prediction, molecular generation, virtual screening, docking, and property prediction.

This creates a potentially powerful feedback loop:

• Scientific data feeds models.
• Models generate predictions.
• Predictions influence experiments.
• Experiments generate new data.
• New data improves future models.

The objective is therefore not simply faster computing. It is faster organisational learning.

AI Agents Could Change How Drug Discovery Scientists Work

BMS is also moving towards what it describes as hybrid intelligence, where AI systems work alongside researchers rather than replacing them.

Its existing AI programmes include automated support for target identification and validation. BMS also uses a predictive approach before laboratory experimentation, allowing computational predictions to influence which molecules progress into experimental work. 

The new infrastructure could expand this model considerably.

AI agents can potentially perform data-intensive tasks, compare evidence, execute computational workflows, and help researchers evaluate competing hypotheses.

NVIDIA’s BioNeMo Agent Toolkit is designed to give agents specialised tools across biology, chemistry, genomics, and drug discovery. 

The core principle of the agentic AI is that drug discovery scientists remain responsible for scientific judgement, while AI handles more of the computational workload surrounding that judgement.

The Real Competitive Asset Is the Drug Discovery Data

The supercomputer will be impressive. But compute alone is unlikely to drive much value.

That’s where the data is needed. And BMS has it.

BMS has decades of proprietary research data, experimental knowledge, and clinical experience. The new infrastructure provides the scale to turn that information into reusable computational intelligence.

That will be important going forward. If every major company can access increasingly powerful AI hardware, competitive advantage may increasingly depend on the quality, integration, governance, and context of proprietary data.

This is consistent with the U.S. Food and Drug Administration’s (FDA’s) emerging approach to AI in drug development. Its 2026 AI principles emphasise data governance, documentation, context of use, risk-based performance assessment, and lifecycle management. 

Pharmatica’s analysis of AI-driven drug discovery examines this wider transition from individual AI applications towards integrated discovery systems.

What BMS’s AI Factory Means for the Future of Drug Discovery

The BMS NVIDIA supercomputer shows how important R&D infrastructure is.

The winning model is not just AI added to drug discovery, but instead drug discovery redesigned around continuously available computational intelligence.

That raises difficult questions around model validation, proprietary data, cybersecurity, intellectual property, reproducibility, and regulatory credibility. The FDA's draft framework for AI supporting regulatory decisions already points towards risk-based assessment of model credibility for specific contexts of use. 

BMS’s investment is therefore building more than simply computing capacity. It’s testing whether a pharmaceutical company can build a consistent AI learning system spanning drug discovery and development.

Pharmatica tracks this convergence of AI, scientific infrastructure, and pharmaceutical strategy, connecting technology adoption with the decisions that determine whether innovation translates into better R&D outcomes.

Pharmatica: Insight. Connection. Impact.

Image