Lilly Built Custom Supercomputer With Blackwell GPUs
The pharmaceutical company deployed the 1,000-GPU LillyPod system to accelerate molecule screening.
Updated on Sept. 22, 2026 in Quantum Computing
Live Poll
Do you feel optimistic about AI and automation improving the future of your career?
Lilly recently completed the assembly of its custom LillyPod supercomputer in just four months. The system utilizes more than 1,000 Nvidia Blackwell graphics processing units to model biological processes and screen potential drug molecules.
Why it matters
Lilly is integrating massive computational power into its pharmaceutical pipeline to optimize drug discovery. This shift signals a broader trend where biological research increasingly relies on digital simulation to manage production and research bottlenecks.
The LillyPod supercomputer integrates more than 1,000 Nvidia Blackwell GPUs. The system enables virtual screening of molecules and testing of manufacturing variables, such as temperature and pressure, before initiating physical wet lab experiments.
The players
Lilly
A global pharmaceutical company focused on medicine development and the integration of high-performance computing in drug manufacturing.
Diogo Rau
The chief information and digital officer at Lilly responsible for the company's digital transformation and computational strategy.
Nvidia
A semiconductor company known for its high-performance GPUs and specialized hardware architectures designed to power AI and large-scale computing tasks.
The details
The LillyPod system functions by creating an AI-assisted digital twin—a virtual, dynamic replica of a physical system—that simulates manufacturing production environments. By testing variables like temperature and pressure within this virtual space, engineers identify and resolve production bottlenecks before physical implementation. This computational modeling allows the company to screen complex biological molecules at scale, reducing the reliance on traditional trial-and-error methods in the lab.
Timeline
2026: Diogo Rau was named Executive of the Year.
Four months: Duration taken to construct the LillyPod supercomputer.
10 years: Estimated timeframe for the reduction of traditional wet lab scientists.
The Tech Race
Lilly's deployment of the Blackwell architecture follows an industry-wide race to integrate generative AI and accelerated computing into the drug discovery pipeline. The firm positions this infrastructure to compete against other life sciences organizations building proprietary supercomputers to optimize lead identification.
The transition to computational drug-making aims to streamline production, potentially leading to faster manufacturing cycles for new medicines. Scientists in the field should anticipate a shift in workflow as physical wet lab tasks are increasingly augmented or replaced by digital simulations.
The takeaway
Lilly is banking on the idea that high-performance compute cycles can effectively replace physical experiments in the early stages of drug development. Watch for the 10-year outlook on workforce composition to see if this computational shift significantly reduces the need for large-scale, traditional wet lab staffing.
Further reading
For more on how high-performance clusters are reshaping research, see our Quantum Computing section.
Source note: This article includes information reported by TIME.
Live Poll
Do you feel optimistic about AI and automation improving the future of your career?
