Now Live: LillyPod AI factory for Pharmaceutical Discovery and Development
Lilly has turned on what it calls the industry’s most powerful AI supercomputer wholly owned and operated by a pharmaceutical company. The LillyPod AI factory is now live, designed to accelerate discovery and development with end-to-end AI workflows and large-scale training and inference on biomedical and chemical data—positioning Eli Lilly at the forefront of an emerging AI factory model in regulated life sciences [1][3][4].
What is LillyPod? scope, purpose, and objectives
LillyPod is a full-stack AI factory built to handle data ingestion, training, fine-tuning, and inference across discovery and development workflows. It is intended to power AI agents for molecular design, simulation, and optimization, as well as digital twins for pharmaceutical manufacturing. Select proprietary models trained on internal experimental data will be made available to external biopharma and biotech partners through Lilly TuneLab, a federated AI/ML platform that enables access to Lilly-trained models without sharing raw data [1][2][3].
Under the hood: hardware and architecture
At the core is the world’s first NVIDIA DGX SuperPOD using DGX B300 systems with 1,016 NVIDIA Blackwell Ultra GPUs, delivering over 9 exaflops (more than 9,000 petaflops) of AI performance. The deployment integrates nearly 5,000 connections with over 1,000 pounds of fiber cabling and was physically assembled in roughly four months—underscoring the pace at which pharma-grade AI infrastructure is coming online [1][6].
A unified high-speed NVIDIA Spectrum-X Ethernet fabric connects GPUs, storage, and systems over a single network, while NVIDIA Mission Control provides orchestration, performance monitoring, and automated AI operations across the SuperPOD. Together, these components form a turnkey AI factory architecture adapted for regulated healthcare use cases [1][6].
Software, orchestration, and operational workflows
The platform uses NVIDIA’s full-stack AI factory design—accelerated computing, Spectrum-X networking, and Mission Control—to automate workload scheduling, monitor performance, and standardize operations across a single, high-speed fabric. This alignment is intended to streamline compliant workflows in life sciences, from data preparation and model training to large-scale inference and evaluation, all within a managed environment [1][6].
Primary use cases: from molecular design to digital twins
Lilly plans to train large biomedical foundation models and deploy AI agents for molecular design, simulation, and optimization, alongside digital twins for pharmaceutical manufacturing. By centralizing compute, networking, and orchestration, the company aims to expand the scale and complexity of experiments and accelerate discovery timelines, while noting that actual results depend on scientific and regulatory realities [1][3][4].
LillyPod AI factory — partner access and federated models
Lilly TuneLab provides a federated access model so partners can tap select Lilly-trained models without sharing raw data—designed to preserve data privacy while extending the reach of AI capabilities across the biopharma ecosystem. For startups and established firms exploring how biotech partners can use federated AI from pharma companies, this approach could catalyze collaborations without centralizing sensitive datasets [1][2][3].
For broader context on the underlying infrastructure, see NVIDIA’s product overview of the DGX SuperPOD with Blackwell Ultra GPUs in its official announcement here (external) [6].
Energy, sustainability, and operations
Lilly targets operating the system on 100% renewable electricity by 2030. Efficient liquid cooling is intended to limit incremental energy impact—an increasingly important consideration for energy efficiency and liquid cooling for AI datacenters. The physical build underscores operational pragmatism, with rapid assembly timelines and dense connectivity integrated into a single network fabric [1].
Risks, limitations, and regulatory context
Public statements are careful to frame benefits as forward-looking expectations subject to scientific, technical, regulatory, and commercial uncertainties. While the AI factory for drug discovery may expand experimental scale and speed, outcomes will ultimately depend on validation, compliance, and market dynamics across the development pipeline [1][3].
What this means for pharma, biotech partners, and investors
- For R&D leaders: The combination of Blackwell Ultra GPUs and Mission Control orchestration targets scalable training and inference, which could open new fronts in large-model experimentation for drug design [1][6].
- For infrastructure teams: The NVIDIA DGX SuperPOD pharma stack—DGX B300 systems, Spectrum-X Ethernet, and unified fabric—offers a blueprint for enterprise AI infrastructure in healthcare [1][6].
- For partners: Lilly TuneLab’s federated approach provides a path to leverage Lilly-trained models without sharing raw data, potentially reducing integration barriers and data-governance friction [1][2][3].
- For investors: The initiative aligns with a broader push to build AI factories across healthcare, including deployments at other institutions, indicating momentum behind specialized, domain-grade AI infrastructure [1][4].
Practical next steps for readers
- Technical teams evaluating NVIDIA DGX SuperPOD pharma architectures can track updates on orchestration, networking, and model throughput as Lilly publishes further details [1][6].
- Partnership leads can explore Lilly TuneLab federated models for potential collaborations and access pathways to select Lilly-trained models [1][2][3].
- Strategy and sustainability leaders should monitor progress toward the 2030 renewable electricity target and the operational profile of liquid cooling at scale [1].
For additional market coverage and implementation insights, you can also Explore AI tools and playbooks.
Hero image: NVIDIA DGX SuperPOD infrastructure powering Eli Lilly’s AI factory [1].
Sources
[1] Now Live: Lilly AI Factory for Pharmaceutical Discovery and Development
https://blogs.nvidia.com/blog/lilly-ai-factory-live/
[2] Lilly Forms Partnership with NVIDIA to Create AI Supercomputer
https://www.pharmexec.com/view/lilly-forms-partnership-nvidia-create-ai-supercomputer
[3] Lilly partners with NVIDIA to build the industry’s most powerful AI supercomputer, supercharging medicine discovery and delivery for patients
https://www.prnewswire.com/news-releases/lilly-partners-with-nvidia-to-build-the-industrys-most-powerful-ai-supercomputer-supercharging-medicine-discovery-and-delivery-for-patients-302597285.html
[4] NVIDIA and Partners Build America’s AI Infrastructure and Create Blueprint to Power the Next Industrial Revolution
https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-and-Partners-Build-Americas-AI-Infrastructure-and-Create-Blueprint-to-Power-the-Next-Industrial-Revolution/default.aspx
[5] Powering Pharma AI with NVIDIA H100 and Blackwell GPUs
https://intuitionlabs.ai/articles/nvidia-gpus-in-pharma-industry
[6] NVIDIA Blackwell Ultra DGX SuperPOD Delivers Out-of-the-Box AI Factories
https://nvidianews.nvidia.com/news/blackwell-ultra-dgx-superpod-supercomputer-ai-factories
