Bristol Myers Squibb Joins Nvidia’s Pharma AI Supercomputer Arms Race

Bristol Myers Squibb said Monday it is building the most powerful AI supercomputer any single drugmaker has assembled, an eight-system Nvidia cluster running on new Vera Rubin chips. It is also the third time in nine months that a major pharmaceutical company has made almost that exact claim.

Eli Lilly made the claim in October. Roche made it in March. Each time, the ‘largest AI supercomputer in life sciences’ title got a little more crowded, and a little harder to verify.

Eight Vera Rubin Systems Join BMS’s Cluster

The new system is a second Nvidia DGX SuperPOD, built from eight rack-scale Vera Rubin NVL72 systems, Bristol Myers Squibb (BMS) and Nvidia said. Together with the company’s original cluster, installed nearly three years ago, the two will merge into one unified environment reachable from every BMS site worldwide.

Nvidia says the combined hardware delivers up to 10 times the performance per megawatt of the system it replaces, pairing new Vera CPUs with Rubin GPUs. Greg Meyers, BMS’s chief digital and technology officer, said the upgrade reflects how quickly scientists filled the space they already had.

“We actually consumed all the space we had,” Meyers told STAT. Erin Davis, BMS’s vice president of research business insights and technology, described the same crunch to Nvidia’s blog. “We’re in production with some very large-scale predictions around large molecules,” she said. “We’re building our own foundational models, and that takes a lot of GPUs.”

“BMS has made a deliberate bet on AI, and we are beginning to see it pay off in our pipeline and operations,” Meyers said in the announcement. The expanded cluster will train foundation models on BMS’s own research data and power agentic workflows across oncology, hematology, cardiovascular disease, immunology and neuroscience, drawing on BioNeMo, Nvidia’s platform for biological AI.

The Third ‘Biggest AI Supercomputer’ Claim Since October

BMS is not the first to reach for the superlative. STAT News counted it as the third drugmaker in nine months to announce it was building the largest AI supercomputer in the life sciences industry.

Each announcement leaned on Nvidia’s newest chips, and each came with its own record-setting language.

Lilly’s system, nicknamed LillyPod, was unveiled at Nvidia’s GTC conference in Washington and packs 1,016 Blackwell Ultra GPUs delivering over 9,000 petaflops of AI performance, assembled in just four months. Roche followed in March by adding 2,176 Nvidia Blackwell GPUs, pushing its total to 3,500 chips across sites in the US and Europe.

How Can Three Drugmakers All Claim to Be Biggest?

Each claim is true because each company defined its own category. Lilly claims the largest system owned and operated by a drugmaker. Roche claims the largest hybrid-cloud AI factory. BMS claims the largest single-owned Nvidia infrastructure in life sciences. None of the three rankings actually overlaps enough to be compared head to head.

• AI factory – Nvidia’s term for a data center built to run the entire AI lifecycle, from ingesting raw data through training, fine-tuning and high-volume inference, all on one connected system.

No independent body benchmarks these claims against each other the way the Top500 list ranks government and academic supercomputers. Each company writes its own definition, then Nvidia publishes it on the corporate blog. That leaves “biggest” as a self-graded exam that three drugmakers have all managed to pass within the same nine months.

A $2.5 Billion Restructuring Runs Alongside the Buildout

The computing buildout is landing in the middle of a multi-year cost-cutting drive at BMS. The company’s 2023 Restructuring Plan is expected to reach $2.5 billion in restructuring charges through 2027, with $1.8 billion already recorded, according to the company’s latest quarterly filing with the Securities and Exchange Commission.

• 2024: BMS unveils a $1.5 billion cost-cutting plan built around roughly 2,220 job losses worldwide.
• February 2026: A second, $2 billion “strategic productivity initiative” arrives, with 247 New Jersey positions cut in its first wave.
• April 2026: A WARN notice adds 206 more New Jersey jobs, split across the second half of the year.
• Since January 2025: Local filings tracked by New Jersey news outlets put cumulative cuts above 1,700 positions.

The pressure traces to patent cliffs. BMS is already feeling the loss of exclusivity on blood cancer drug Revlimid, and faces the same fate soon for cancer immunotherapy Opdivo and blood thinner Eliquis, which is also among the first medicines subject to Medicare price negotiation under the Inflation Reduction Act.

Nvidia and BMS have not disclosed what the new Vera Rubin cluster costs. Lilly’s buildout came bundled with a separate $1 billion, five-year research lab it agreed to build with Nvidia in January, the closest public benchmark for what this scale of computing now costs a drugmaker.

Predict First, and the Judgment Nvidia Can’t Replace

Robert Plenge, BMS’s executive vice president and chief research officer, framed the investment as a way to sharpen decisions, not just speed them up.

Drug discovery is a sequence of decisions made under uncertainty, and better decisions come from better evidence, faster.

Plenge said in the announcement. “The goal isn’t speed for its own sake; it’s raising the probability that each program we advance is the right one,” he added. BMS describes the approach internally as “Predict First,” using AI-generated predictions to guide experimental design before lab work begins.

Davis has her own nickname for the setup: the “SuperDuperPOD.” When Meyers asked her whether she was sure she could even fill it with work, her answer left no room for doubt. “Just give us time,” she told him, according to Nvidia.

The plan is to open the machine to every scientist at BMS rather than a select few. “Instead of equipping a small group of researchers with access to the supercomputer, we’re opening it up to literally every scientist,” Davis said. “No one has to wait, and no one is told they have a limit.”

Nvidia Wins Regardless of Who Claims Biggest

Whatever else these announcements prove, they all point to the same buyer. Every “largest AI supercomputer” claim in pharma this year has run on Nvidia hardware, and every drugmaker chasing the title is spending more with the same chipmaker to get there.

That dependence has started to raise its own questions. When Nvidia agreed in January to help fund the $1 billion, five-year research lab with Lilly, Reuters reported that neither company would say whether Nvidia’s money would eventually flow back to Nvidia through chip purchases, a circular arrangement that has drawn scrutiny around other Nvidia-funded deals.

Lilly’s chief information and digital officer, Diogo Rau, has described the ambition in blunter terms, saying the goal is to escape what he calls the traditional pharma industry life cycle.

So far, none of the three drugmakers claiming the industry’s biggest AI cluster has published data showing the hardware shortened a single drug’s path to approval. The chips are running. The comparison to actual medicines is still pending.

Frequently Asked Questions

What is an Nvidia DGX SuperPOD?

It’s Nvidia’s prepackaged supercomputer architecture: racks of GPUs and CPUs wired together on one high-speed network so they function as a single machine rather than separate servers. BMS, Lilly and Roche have each built theirs around different chip generations, from Blackwell to the newer Vera Rubin line.

How much is Bristol Myers Squibb spending on its Nvidia buildout?

BMS and Nvidia have not disclosed a price for the new cluster. The closest public benchmark comes from Lilly, which agreed in January to spend up to $1 billion over five years with Nvidia on a related Bay Area research lab.

What other pharma companies have signed big AI infrastructure deals?

Merck struck a deal worth up to $1 billion with Google Cloud for an agentic AI system spanning its R&D, manufacturing and commercial operations. Novo Nordisk partnered with OpenAI, and Sanofi put $294 million into an AI hub in Toronto.

Has AI actually sped up drug discovery at Bristol Myers Squibb?

BMS says its AI agents already automate target identification and validation, saving scientists weeks of manual work, and that its models produce large-scale predictions on how big molecules behave. The company has not published independent data tying the technology to a faster drug approval.

What is Nvidia’s Vera Rubin platform?

Vera Rubin is Nvidia’s newest generation of AI chips, pairing new Vera CPUs with Rubin GPUs. Nvidia says systems built on the platform deliver up to 10 times more performance per megawatt than the Blackwell-based systems that came before it.