NVIDIA invests 1 billion USD to collaborate with Lilly to build AI drug discovery labs, accelerating drug development automation

Nvidia (Nvidia) announced that it will invest $1 billion over the next five years to jointly build an AI drug discovery laboratory in Silicon Valley with the American pharmaceutical company Eli Lilly (Eli Lilly). The goal is to truly bring AI into the drug R&D process and accelerate workflows that traditionally rely heavily on manpower and physical experiments.

The laboratory is located in San Francisco, where scientists and AI engineers work side by side.

This AI co-innovation laboratory will be based in the San Francisco Bay Area and operate under a “co-location” model, allowing Lilly’s biologists, scientists, and medical experts to work in the same space as Nvidia’s AI model engineers and system engineers.

Both parties hope to generate high-quality data through face-to-face collaboration while training and fine-tuning biological and chemical models on Nvidia’s BioNeMo platform, speeding up the new drug development process.

(Note: BioNeMo is an open machine learning framework designed specifically for deep learning models in biopharmaceuticals. It supports biological molecule research using DNA, RNA, and protein data. It offers optimized training workflows, data tools, and pre-trained model architectures to accelerate the most time-consuming and expensive phase of model development, enabling researchers to build high-performance biomedical AI models faster and at larger scales.)

Simulate with computers first, then conduct molecular experiments in the lab.

Nvidia CEO Jensen Huang stated that scientists can first simulate and explore biological and chemical spaces extensively on computers, screening for the most promising molecules before moving on to physical experiments.

Lilly CEO David Ricks pointed out that with nearly 150 years of scientific knowledge and data, combining Nvidia’s computational power and model-building techniques could fundamentally transform drug R&D.

Building a continuous learning system with two types of laboratories providing 24-hour mutual feedback.

In the initial phase of collaboration, the focus will be on establishing a “continuous learning system” that tightly links Lilly’s physical laboratories with computer simulation and modeling laboratories.

Experimental results will be fed back to AI models in real-time, which will then suggest new experiments. This cycle will continue 24/7, continuously optimizing experiments, data, and models.

Using Nvidia’s BioNeMo and Vera Rubin architecture to train AI biomedical models.

Technologically, the laboratory will use BioNeMo as the core platform to train next-generation biological and chemical foundational and frontier models. The computational architecture will also incorporate the next-generation Vera Rubin to support the massive training demands of these models.

Lilly’s previously announced AI factory will also serve as the main training base for these models, focusing on large-scale biomedical model training tasks. This collaboration extends beyond drug discovery. Both parties will explore applying AI to clinical trials, manufacturing, and commercial operations, integrating multimodal models, agent-based AI, robotics, and digital twins.

(Note: Vera Rubin is Nvidia’s next-generation AI computing platform architecture, officially mass-produced in 2026. As the successor to Blackwell, it consists of the high-performance Vera CPU emphasizing extreme efficiency and the Rubin GPU, which uses HBM4 memory and TSMC’s 3nm process. Its goal is to drive a new era of large-scale agent-based AI and physical AI worldwide through a thousandfold increase in computing power and one-tenth of the cost.)

Robots and physical AI advancing factories to enhance production capacity and supply chain stability.

On the manufacturing side, Lilly will introduce physical AI and robots to strengthen the production capacity of high-demand drugs.

Using Nvidia’s Omniverse and RTX PRO servers, Lilly can first create digital twins of entire manufacturing lines and supply chains in virtual environments for stress testing and optimization, then implement in real factories to reduce risks and costs.

(Note: Omniverse is Nvidia’s virtual world operating system, and RTX PRO servers are the most powerful hardware engines running this system. In enterprise applications by 2026, these are often bundled together to create digital twins and train physical AI, such as robots.)

Connecting startups and research communities to expand the biomedical AI ecosystem.

Nvidia emphasizes its role in open-source AI, providing models, data, and tools to help enterprises develop practical AI systems. Additionally, Nvidia’s Inception program offers startups technical guidance, software, and computing resources.

Lilly’s TuneLab platform allows biotech companies to utilize Lilly’s years of accumulated drug discovery models. In the future, Nvidia’s Clara life sciences open models will also be integrated into new workflows.

This co-innovation laboratory will become an important resource hub for Nvidia and Lilly’s startup ecosystem and researchers.

(Siemens and Nvidia jointly promote industrial AI systems: From digital twins to autonomous factories, accelerating AI adoption in manufacturing.)

This article about Nvidia investing $1 billion to partner with Lilly to build an AI drug discovery lab and accelerate drug development automation first appeared on Chain News ABMedia.