Eli Lilly has officially brought "LillyPod" online, a massive AI factory designed to revolutionize how medicines are discovered and developed. Built in just four months, this computational powerhouse features over 1,000 NVIDIA Blackwell Ultra GPUs, providing the sheer muscle needed to process 700 terabytes of genomic data. By shifting focus from traditional physical experiments to high-scale digital simulations, the company aims to "break the physical limit" of the laboratory.

The system isn't just about raw speed; it's a sophisticated instrument for training advanced AI architectures. Lilly plans to use the cluster for protein diffusion models (which help design new biological structures) and graph neural networks tailored for small-molecule discovery. This allows scientists to test billions of molecular hypotheses virtually before ever touching a test tube, significantly reducing the time and cost of early-stage research.

To democratize these advancements, Lilly is launching TuneLab, a platform that grants biotech partners access to proprietary drug discovery models. Using a privacy-preserving technique called federated learning, different companies can contribute to model training without sharing their sensitive underlying data. This collaborative approach, powered by NVIDIA's BioNeMo and FLARE frameworks, signals a shift toward an interconnected, AI-driven ecosystem for global health.

Eli Lilly has officially brought "LillyPod" online, a massive AI factory designed to revolutionize how medicines are discovered and developed. Built in just four months, this computational powerhouse features over 1,000 NVIDIA Blackwell Ultra GPUs, providing the sheer muscle needed to process 700 terabytes of genomic data. By shifting focus from traditional physical experiments to high-scale digital simulations, the company aims to "break the physical limit" of the laboratory.

The system isn't just about raw speed; it's a sophisticated instrument for training advanced AI architectures. Lilly plans to use the cluster for protein diffusion models (which help design new biological structures) and graph neural networks tailored for small-molecule discovery. This allows scientists to test billions of molecular hypotheses virtually before ever touching a test tube, significantly reducing the time and cost of early-stage research.

To democratize these advancements, Lilly is launching TuneLab, a platform that grants biotech partners access to proprietary drug discovery models. Using a privacy-preserving technique called federated learning, different companies can contribute to model training without sharing their sensitive underlying data. This collaborative approach, powered by NVIDIA's BioNeMo and FLARE frameworks, signals a shift toward an interconnected, AI-driven ecosystem for global health.