Bristol Myers Squibb Acquires Nvidia AI Supercomputer to Accelerate Drug Discovery

 

AI SUPERCOMPUTER FOR NEW MEDICINES

Global pharmaceutical company Bristol Myers Squibb is investing in Nvidia’s next-generation AI computing technology to speed up the discovery and development of new medicines.

Bristol Myers Squibb, or BMS, announced that it is purchasing an Nvidia DGX SuperPOD powered by the chipmaker’s new Vera Rubin architecture.

The company said it will become the first life sciences organization to acquire a DGX SuperPOD based on Vera Rubin, Nvidia’s next-generation AI computing platform.

The new system will be used across BMS research operations to train proprietary artificial intelligence models, analyze scientific data, predict molecular behavior, and identify promising drug candidates before they undergo laboratory testing.

Expanding AI Computing for Medical Research

The new computing cluster will include eight DGX Vera Rubin NVL72 systems. Each rack-scale system combines Nvidia Vera central processing units with Rubin graphics processing units designed for demanding AI workloads.

BMS already operates an older Nvidia DGX SuperPOD that has been in use for approximately three years. Company executives said the existing infrastructure is now operating at full capacity and is already several generations behind the Vera Rubin platform.

Once deployed, the two SuperPOD systems will operate within a shared computing environment that can be accessed by BMS scientists from research facilities around the world.

The infrastructure will support AI training, scientific predictions, software development, clinical research, small and large-molecule design, and potential digital-twin applications.

Financial details of the acquisition were not disclosed. BMS also did not reveal the exact deployment date or location of the new system.

AI Across Drug Discovery Programs

BMS said artificial intelligence is already helping guide the design of every small-molecule research program and most of its large-molecule programs.

The technology is being used for:

• Identifying potential drug targets
• Optimizing promising compounds
• Predicting protein and molecular behavior
• Developing internal foundation models
• Screening possible medicines before laboratory testing
• Accelerating the production of treatments for clinical trials

According to BMS, AI-supported target identification has already reduced some manual research processes by several weeks.

Robert Plenge, chief research officer of BMS, said the new computing system will allow researchers to evaluate significantly more possible drug candidates during the early stages of development.

Instead of examining only a small number of compounds, scientists could use AI to assess dozens of possible candidates before selecting the most promising ones for laboratory testing.

Predict First Before Laboratory Testing

BMS applies an approach called “Predict First,” which uses AI-generated predictions to eliminate molecules that are unlikely to meet a drug program’s requirements.

Researchers can then prioritize compounds with the best predicted combination of safety, effectiveness, stability, and other essential properties.

Payal Sheth, senior vice president of therapeutic discovery sciences at BMS, said the system helps ensure that valuable laboratory experiments are focused on molecules with the highest probability of success.

This approach does not replace laboratory testing. Instead, it reduces the number of weak candidates entering the laboratory and helps researchers use their time, resources, and equipment more efficiently.

Human scientists will continue to review AI-generated results and make the final decisions on which compounds and research programs should move forward.

Supporting Cancer and Sickle Cell Research

BMS has also used AI to expand its library of CELMoD compounds, which are designed to selectively destroy proteins associated with cancer.

The company is studying these compounds for blood cancers and other diseases. AI modelling has helped researchers examine more protein targets and possible compounds before deciding which candidates should undergo experimental testing.

Plenge also cited an experimental treatment for sickle cell disease as an example of AI-supported research. He said the candidate probably would not have been discovered without the company’s AI tools.

BMS clarified that these improvements refer to the time required to discover and produce potential medicines for clinical testing. They do not guarantee that the treatments will succeed during clinical trials.

The company said AI has already reduced some medicine development processes by approximately 20 to 30 percent. BMS believes the reduction could eventually reach 50 percent.

Nvidia BioNeMo for Biological Research

The Vera Rubin system will give BMS scientists access to Nvidia’s BioNeMo Agent Toolkit, a collection of AI tools developed for biology and drug discovery.

BioNeMo supports scientific applications such as:

• Protein structure prediction
• Molecular generation
• Molecular docking
• Genomics and sequence analysis
• Biological data processing
• Integration of multiple research tools into one workflow

BMS is also introducing natural-language tools that will allow more researchers to initiate complex prediction tasks without requiring advanced knowledge of computing systems.

This could make powerful AI research tools accessible not only to computational specialists, but also to a broader group of scientists across the organization.

Connecting Scientists Around the World

The expanded infrastructure will connect BMS research locations through a common data and computing environment.

For example, scientific data produced by researchers in Lawrenceville, New Jersey, could be incorporated into models used by scientists in San Diego and other locations.

The shared platform will retain information from experiments, clinical results, and external research partnerships. This will help ensure that discoveries and lessons from one research team can be accessed and applied by other teams.

The environment will be managed using Nvidia Mission Control, which provides cluster provisioning, infrastructure monitoring, and workload management.

More AI Power Using Less Electricity

BMS and Nvidia said the eight-system Vera Rubin cluster could deliver up to 10 times more computing performance per megawatt than the infrastructure it is expected to replace.

Greg Meyers, chief digital and technology officer of BMS, said energy efficiency is becoming increasingly important as pharmaceutical companies deploy larger AI models and more demanding scientific workloads.

The investment reflects a growing shift in the pharmaceutical industry, where AI is becoming a central research tool rather than a limited experimental technology.

By combining Nvidia’s next-generation computing infrastructure with human scientific expertise, Bristol Myers Squibb hopes to examine more possible treatments, reduce unnecessary laboratory work, and bring promising medicines into clinical development faster.