Bristol Myers Squibb Builds the Life Sciences Industry’s Most Advanced AI Factory with NVIDIA Vera Rubin
Bristol Myers Squibb (BMS) is taking a major step toward transforming the future of drug discovery by deploying what it describes as the life sciences industry’s most advanced AI factory, powered by NVIDIA’s next-generation DGX Vera Rubin platform. The investment reflects the pharmaceutical company’s growing confidence in artificial intelligence as a core driver of scientific innovation, enabling researchers to accelerate discoveries, improve decision-making, and bring new therapies to patients faster.
At the Bristol heart of the initiative is a second NVIDIA DGX SuperPOD, built using eight DGX Vera Rubin NVL72 systems. Together, these rack-scale AI systems form one of the most powerful computing environments ever deployed in pharmaceutical research. Designed to deliver up to 10 times greater performance per megawatt than the infrastructure it replaces, the new AI cluster significantly expands Bristol Myers Squibb’s computational capabilities while improving energy efficiency.
For the company’s scientists, however, the true value lies not in the hardware itself but in what it enables.
Democratizing AI Across Research
Erin Davis, Vice President of Research Business Insights and Technology at Bristol Myers Squibb, affectionately refers to the new deployment as the “SuperDuperPOD.” The nickname reflects both the scale of the system and the company’s ambitious vision for how AI should be used throughout its research organization.
Rather than limiting access to a select group of computational experts, Bristol Myers Squibb plans to make high-performance AI resources available to every scientist across the company.
Instead of waiting in computing queues or competing for limited GPU allocations, researchers will have immediate access to the computing power needed to train AI models, analyze biological data, simulate molecular interactions, and perform large-scale predictions.
This shift fundamentally changes how scientists interact with AI. Instead of worrying about infrastructure limitations, they can focus their attention on solving complex biological problems and discovering new medicines.
The Bristol new AI factory will provide researchers with a unified platform supporting advanced AI workflows, including NVIDIA BioNeMo Agent Toolkit for biological AI, enabling everything from predictive modeling to autonomous research assistants that support drug discovery.
Accelerating Drug Discovery
Artificial intelligence has already become an integral part of Bristol Myers Squibb’s research strategy.
The company has operated its first NVIDIA DGX SuperPOD for approximately three years, using it to streamline multiple stages of the drug discovery process.
One of the most significant improvements has been AI-assisted target identification, where machine learning models rapidly analyze biological data to identify promising therapeutic targets.
Tasks that once required weeks of manual analysis can now be completed in a fraction of the time, allowing researchers to spend more time interpreting results and designing experiments instead of processing data.
The company’s AI capabilities have also contributed to expanding its library of CELMoD compounds—engineered molecules designed to selectively degrade disease-causing proteins.
Originally developed for blood cancer treatments, these compounds are now being explored across a broader range of diseases, opening new possibilities for future therapies.
By analyzing vast chemical spaces and identifying promising molecular candidates more efficiently, AI enables scientists to investigate opportunities that would have been impractical using conventional research methods.
Predict First: Smarter Experimental Design
Bristol Myers Squibb is also using AI to transform lead optimization through an approach known internally as “Predict First.”
Rather than synthesizing large numbers of molecules and evaluating them experimentally, researchers first use AI models to predict which compounds are most likely to possess the desired biological and chemical properties.
These predictions help prioritize laboratory work, ensuring that experimental resources are focused on molecules with the highest probability of success.
According to Payal Sheth, Senior Vice President of Therapeutic Discovery Sciences, predictive modeling allows scientists to optimize multiple molecular characteristics simultaneously before entering the laboratory.
By filtering out compounds unlikely to meet development criteria, researchers reduce unnecessary experimentation while accelerating the progression of promising drug candidates.
This predictive approach not only shortens research timelines but also enables more efficient use of laboratory resources.
Scaling AI Requires More Compute
As AI becomes increasingly embedded across research workflows, computational demand has grown dramatically.
Large molecular modeling projects, foundation model development, and predictive biological simulations all require enormous GPU resources.
According to Davis, existing infrastructure has reached its limits.
The company’s researchers are already running large-scale prediction models while simultaneously building proprietary foundation models tailored specifically for pharmaceutical research.
Meeting these growing demands requires infrastructure capable of supporting continuous AI innovation without creating bottlenecks for scientists.
The new DGX Vera Rubin-powered SuperPOD provides that foundation, offering enough computing capacity to support expanding AI workloads across every stage of drug discovery.
From Computational Chemist to Technology Leader
Davis brings a unique perspective to this transformation.
Originally trained as a computational chemist, she spent years conducting scientific research before realizing that available computing technologies were limiting what researchers could accomplish.
Rather than accepting those limitations, she shifted her career toward building better research technologies.
Over the following fifteen years, Davis worked with companies including ChemAxon, Schrödinger, and X-Chem, helping develop enterprise software platforms for pharmaceutical research.
Across every organization, she observed the same challenge.
The problem was rarely the underlying technology.
Instead, it was ensuring that powerful computational tools became accessible to the scientists who needed them most.
That philosophy now shapes Bristol Myers Squibb’s AI strategy.
Personal Motivation Behind Innovation
For Davis, accelerating drug discovery is more than a professional objective.
Her father passed away after battling Alzheimer’s disease, an experience that deeply influenced her commitment to advancing biomedical research.
Although no cure was available, she believes even therapies capable of slowing symptoms could have significantly improved both his quality of life and the well-being of their family.
Diseases affecting brain health remain among medicine’s greatest challenges, requiring deeper biological understanding and more effective therapeutic discovery methods.
AI represents one of the most promising tools available to accelerate progress in these difficult research areas.
Building a Connected Global Research Network
The new AI infrastructure will not operate independently.
Instead, Bristol Myers Squibb plans to integrate its existing DGX SuperPOD with the new DGX Vera Rubin deployment into a single global computing environment.
Researchers across every Bristol Myers Squibb location will access the same shared data platform regardless of geographic location.
Previous barriers created by separate research sites, legacy infrastructure, or specialized computing requirements will be replaced by AI-native workflows managed through NVIDIA Mission Control.
Scientists will increasingly interact with complex AI systems using natural language rather than specialized programming knowledge.
This democratization of computing enables researchers worldwide to collaborate more effectively while sharing data, models, and scientific insights across therapeutic programs.
Creating Institutional Intelligence
According to Sheth, one of AI’s greatest contributions extends beyond accelerating individual research projects.
It enables cumulative organizational learning.
Historically, many pharmaceutical programs operated largely independently, with lessons learned remaining isolated within individual teams.
Today’s AI systems create continuous learning loops where knowledge generated from one research program strengthens future programs across the organization.
Experimental results from laboratories in New Jersey, for example, can immediately improve predictive models used by scientists in California or elsewhere around the world.
Each experiment, clinical outcome, and research collaboration contributes to a growing body of institutional intelligence that benefits every subsequent discovery effort.
The Rise of Agentic AI
Looking ahead, Bristol Myers Squibb sees enormous potential in agentic AI—AI systems capable of autonomously assisting scientists throughout complex research workflows.
Unlike traditional software that performs isolated tasks, AI agents can operate across multiple research domains, connecting information that might otherwise remain fragmented.
These agents can analyze experimental data, review scientific literature, recommend next steps, and coordinate information across therapeutic programs.
Davis believes this capability fundamentally changes pharmaceutical research.
Rather than working with isolated computational tools, scientists will collaborate with teams of AI assistants trained using Bristol Myers Squibb’s accumulated scientific knowledge.
Importantly, neither Davis nor Sheth views AI as replacing scientific expertise.
Instead, they emphasize that AI augments human judgment by providing richer predictions, broader analysis, and faster access to relevant knowledge.
Human researchers remain responsible for asking critical questions, evaluating evidence, interpreting uncertainty, and guiding scientific strategy.
AI simply enables them to accomplish far more than previously possible.
An AI Factory Built for Every Stage of Discovery
The new DGX Vera Rubin infrastructure has already been mapped to support virtually every phase of Bristol Myers Squibb’s research pipeline.
Planned workloads include small-molecule discovery, biologics research, clinical development, digital twin technologies, and advanced predictive modeling.
Rather than building the world’s most powerful AI cluster for its own sake, the company has designed the platform to generate measurable scientific impact across every stage of therapeutic development.
As AI continues reshaping pharmaceutical research, Bristol Myers Squibb’s investment represents more than an infrastructure upgrade—it signals a broader transformation in how medicines are discovered.
By combining cutting-edge computing, advanced AI models, biological expertise, and collaborative research workflows, the company aims to accelerate scientific breakthroughs while giving every researcher access to the computational resources needed to solve some of medicine’s most challenging problems.
Source link: https://blogs.nvidia.com