Lilly and NVIDIA build 1,000-GPU AI supercomputer for drug discovery

Eli Lilly said on Oct. 28, 2025 that it is building a supercomputer with NVIDIA that it describes as the most powerful owned and operated by a pharmaceutical company. The machine is the world's first NVIDIA DGX SuperPOD built with DGX B300 systems and runs on more than 1,000 B300 GPUs connected by one high-speed network ^[1].

Lilly says the system will support an "AI factory," which it describes as computing infrastructure that handles the whole AI lifecycle, from taking in data and training models to fine-tuning and running them at high volume. The company says scientists will be able to train AI models on millions of experiments to test potential medicines ^[1].

Some of those models will be shared on Lilly TuneLab, a collaborative drug discovery platform Lilly created to widen access to advanced discovery tools across the biopharma industry. Lilly says TuneLab will add workflows that use selected NVIDIA Clara open-source models ^[1].

Beyond discovery, Lilly says it plans to use the supercomputer to shorten development cycles, run "scientific AI agents" that support reasoning and planning, improve medical imaging and biomarker development, and use digital twins with NVIDIA robotics to improve manufacturing efficiency and reduce downtime ^[1].

"As a 150-year-old medicine company, one of our most powerful assets is decades of data," said Diogo Rau, Lilly's executive vice president and chief information and digital officer. "With purpose-built AI models and AI, we can set a new scientific standard that accelerates innovation to deliver medicines to more patients, faster" ^[1].

Kimberly Powell, NVIDIA's vice president of health care, said AI factories are "becoming the new instrument of science — enabling the shift from trial-and-error discovery to a more intentional design of medicines" ^[1]. Thomas Fuchs, Lilly's chief AI officer, said the company is "shifting from using AI as a tool to embracing it as a scientific collaborator" ^[1].

Lilly said the supercomputer will run on 100 percent renewable electricity inside existing Lilly facilities and use the company's existing chilled water infrastructure for liquid cooling, in line with its commitment to carbon neutrality by 2030 ^[1].

At an event marking Lilly's 150th anniversary, CEO Dave Ricks said the supercomputer was built in Indianapolis and is already helping scientists solve complex problems and is being used in manufacturing. Ricks said Lilly is working with Purdue University on automation and robotics for manufacturing, with a goal of using AI and robotics for unsafe or repetitive jobs ^[2].

Ricks also said Lilly is expanding research into neurodegenerative diseases, including ALS, frontotemporal dementia and Lewy body dementia, and that the company has medicines in clinical trials and other molecules in development aimed at reducing side effects. "We know these medicines can affect addiction and other brain conditions," he said ^[2].

On the GLP-1 market, Ricks said Lilly is developing more powerful medicines to help people reach healthier weights faster and more safely, and highlighted the company's newly launched oral GLP-1 medicine, orforglipron ^[2].

Why it matters for patients

Nothing in this announcement changes the medicines people take today, their doses, or their insurance coverage. The supercomputer is research and manufacturing infrastructure, not a treatment, and Lilly's own forward-looking statement says there is no guarantee the company will realize the expected benefits of the supercomputer or the AI factory, that the systems will achieve the results described, or that they will produce commercially successful products ^[1].

Drug discovery claims of this kind usually take years to show up as approved medicines, if they do at all. Lilly has not disclosed any specific drug candidate that came out of the supercomputer, any timeline for one, or what it cost to build.

For people using GLP-1 medicines, the more concrete item is Ricks' statement that Lilly is working on more powerful weight-management medicines, a claim that came with no data, no trial results and no timeline in the reports of his remarks ^[2]. He also referred to orforglipron as newly launched ^[2].

One detail with longer-range implications is TuneLab: by putting some Lilly and NVIDIA models on a platform open to other biopharma companies, Lilly says it is widening access to discovery tools beyond its own walls ^[1].

What happens next

Lilly said its presentation, "Enterprise-Scale AI for Drug Discovery: Strategy, Infrastructure and Outcomes," took place at NVIDIA's GTC conference in Washington, D.C., on Oct. 28, 2025 ^[1]. The company has not disclosed when the supercomputer will be fully operational, how much it is spending, or which molecules it will work on first ^[1]. No dates have been reported for the brain-disease trials Ricks mentioned or for the more powerful weight-management medicines he described ^[2].