- Bristol Myers Squibb announced July 20 it will deploy an NVIDIA DGX SuperPOD with DGX Vera Rubin NVL72 systems, which the company says will be the most powerful and energy-efficient single-owned NVIDIA infrastructure in life sciences, roughly 15 times more powerful than the supercomputing network the two companies brought online in 2024, according to BMS Chief Digital and Technology Officer Greg Meyers in an interview with Endpoints News.
- BMS Chief Research Officer Robert Plenge said the company has already cut drug development timelines by 20% to 30% using AI tools; the new infrastructure will train proprietary foundation models on decades of BMS data across oncology, hematology, cardiovascular, immunology, and neuroscience.
Bristol Myers Squibb is expanding its nearly three-year partnership with NVIDIA to deploy what it calls the most powerful AI infrastructure in life sciences, the company announced July 20. This marks the third time in nine months a major drugmaker has made a version of that claim.
The new DGX SuperPOD, built on NVIDIA’s DGX Vera Rubin NVL72 systems, will be roughly 15 times more powerful than the supercomputing network BMS and NVIDIA brought online in 2024, Chief Digital and Technology Officer Greg Meyers told Endpoints News. BMS will be the first pharma company to use NVIDIA’s newest chips, and the buildout will be co-located at a commercial data center rather than at company headquarters. Financial terms weren’t disclosed.
The pitch from BMS leadership is deliberately not about raw speed. “The goal isn’t speed for its own sake; it’s raising the probability that each program we advance is the right one,” Chief Research Officer Robert Plenge said in the company’s release. That said, the speed numbers are notable: Plenge said that AI tools have already shortened BMS’s drug development process by 20% to 30%. And according to the company’s announcement, its “Predict First” methodology, where AI-generated predictions shape experimental design before bench work begins, now informs every small molecule program and the majority of large molecule programs.
The Vera Rubin architecture delivers up to ten times the performance per megawatt of its predecessor, according to BMS, letting the company scale AI workloads without a proportional jump in energy consumption. The cluster will train next-generation foundation models on decades of proprietary BMS data and, where useful, draw on BioNeMo, NVIDIA’s biological AI platform. It underpins what BMS calls “hybrid intelligence” — AI co-scientists handling data-intensive execution while human researchers focus on direction and judgment. “Historically, the staples for us have been biology and chemistry,” Meyers told Endpoints. “I think computer science is now an equal third leg of scientific discovery.”
The announcement lands in a crowded field of superlatives. As STAT News health tech reporter Brittany Trang noted, BMS is the third pharma company in nine months to declare it’s building the industry’s largest NVIDIA AI supercomputer. Eli Lilly committed up to $1 billion with NVIDIA for an AI co-innovation lab, per Fierce Biotech, while Roche announced in March what it called pharma’s largest GPU deployment — more than 3,500 NVIDIA Blackwell GPUs spanning discovery, manufacturing, and diagnostics. Novo Nordisk, meanwhile, went a different route in April with a broad OpenAI partnership covering R&D, manufacturing, and supply chain. For BMS, the NVIDIA deal sits alongside a recent enterprise-wide agreement with Anthropic, which Meyers described on LinkedIn as “a shared, single pane of glass” connecting the company’s scattered data and systems, with NVIDIA supplying the computational horsepower underneath.
