Nvidia and Eli Lilly commit up to $1B to AI drug discovery lab

Introduction

Nvidia and Eli Lilly have announced a joint commitment to invest up to $1 billion in an AI co-innovation lab in the San Francisco Bay Area. The goal of this partnership is to use AI to change how new medicines are discovered and how biomedical research gets done.

What the Money is Supposed to Buy

The core idea behind this partnership is proximity. Teams from both companies will work in the same space instead of trading emails across organizational lines. Those co-located teams plan to generate large datasets. They will use that data to train foundation models for biology and chemistry. A foundation model is a large AI system trained on broad data and then adapted for specific jobs. ChatGPT learned language. These models are meant to learn molecules, proteins and the chemistry that connects them.

The Training Process

The training will run through Nvidia's BioNeMo platform. It will also use Nvidia's upcoming Vera Rubin architecture, the chip design that follows the company's current generation of hardware. The lab also plans to bring robotics and what Nvidia calls physical AI into the mix. Those tools are aimed at three areas: experimentation, manufacturing and supply chain operations.

A Bet on AI's Power

Nvidia CEO Jensen Huang and Eli Lilly Chair and CEO David Ricks both framed the venture as a bet on AI's power to transform life sciences. Their pitch combines three ingredients: data, computing power and domain expertise.

Building on Earlier Work

This partnership builds on earlier work. In 2025, the two companies set up a supercomputer for Lilly featuring over 1,000 Grace Blackwell GPUs. The Lilly deal also isn't Nvidia's only science play in the US. The company is working with the US Department of Energy on the Genesis Mission, a public-private initiative launched through an executive order in late 2025.

What This Means for Nvidia, Lilly and AI-Driven Science

For Nvidia, the deal pushes the company further up the value chain. Selling chips to whoever needs them is a fine business. Embedding engineers inside a pharmaceutical giant's research process is a stickier one. If BioNeMo and Vera Rubin become the default tools for training biology and chemistry models at a company of Lilly's scale, that sets a reference design. The 24/7 continuous learning setup is the most ambitious piece. If models can design experiments, robots can run them and results can flow straight back into training, the research cycle could compress.