Now Live: The World’s Most Powerful AI Factory for Pharmaceutical Discovery and Development

2026-02-26 · Source: NVIDIA Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Data Science & Analytics · Depth: Advanced, short

Summary

Eli Lilly and Company has launched LillyPod, the world's most powerful AI factory wholly owned and operated by a pharmaceutical company, built with over 1,000 NVIDIA Blackwell Ultra GPUs. This NVIDIA DGX SuperPOD with DGX B300 systems delivers more than 9,000 petaflops of AI performance and was assembled in just four months. LillyPod enables large-scale training of protein diffusion models, small-molecule graph neural networks, and genomics foundation models, harnessing 700 terabytes of data with over 290 terabytes of high-bandwidth GPU memory. The infrastructure aims to accelerate drug discovery and development by breaking the physical limits of traditional wet lab experiments, allowing scientists to simulate billions of molecular hypotheses computationally.

Key takeaway

For research scientists in pharmaceutical R&D, LillyPod's computational dry lab capabilities fundamentally change drug discovery by enabling the simulation and evaluation of billions of molecular hypotheses. You can now explore vastly more chemical possibilities and apply AI across clinical development and manufacturing, accelerating decision-making and optimizing production in ways previously constrained by physical lab limits.

Key insights

LillyPod, an NVIDIA DGX SuperPOD, significantly accelerates pharmaceutical discovery through massive AI computational power.

Principles

• Computation is central to modern biology and science.
• AI factories enable unprecedented scale in drug discovery.

Method

LillyPod uses NVIDIA DGX SuperPOD with Blackwell Ultra GPUs, NVIDIA Spectrum-X Ethernet, and optimized AI software, managed by NVIDIA Mission Control, to train large-scale protein diffusion, graph neural network, and genomics foundation models.

In practice

• Train protein diffusion models at scale.
• Develop small-molecule graph neural networks.
• Utilize genomics foundation models.

Topics

• AI Supercomputing
• Pharmaceutical Discovery
• NVIDIA Blackwell GPUs
• Foundation Models
• Federated Learning

Best for: Research Scientist, AI Engineer, AI Scientist, Director of AI/ML

Related on AIssential

• Scaling Scientific R&D with AI Supercomputing Infrastructure — with Thomas Fuchs of Eli Lilly — AI supercomputing transforms pharmaceutical R&D by enabling larger models, leveraging vast data, and accelerating scientific discovery and manufacturing.
• Eli Lilly's LillyPod - 9,000 Petaflops of AI Power Just Went Live for Drug Discovery — LillyPod's deployment signals AI's readiness to meet stringent pharmaceutical regulatory standards and accelerate drug discovery.
• BIO Conference with Kimberly Powell — NVIDIA's Bioneo agent toolkit enables "agentic science" to dramatically accelerate drug discovery and scientific research.
• AI model used to generate complete models of proteins in motion — LD-FPG is the first AI framework to generate complete, all-atom dynamic protein models, capturing full motion.
• Introducing new capabilities to GPT-Rosalind — GPT-Rosalind integrates advanced AI with scientific workflows, significantly improving life sciences research efficiency and accuracy.
• Dunia Innovations unveils €280M Berlin GigaLab to industrialise AI-driven materials discovery — The bottleneck in AI-driven materials discovery has shifted from candidate generation to industrial-scale experimental verification.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by NVIDIA Blog.