Nvidia and Eli Lilly Launch $1B AI Drug Discovery Lab
"Combining Lilly's volumes of data and scientific knowledge with Nvidia's computational power could reinvent drug discovery as we know it"
On January 12, 2026, Nvidia and Eli Lilly announced that they're building the world's first artificial intelligence factory for drug discovery. Nvidia and Eli Lilly will invest up to $1 billion over five years in a co-innovation laboratory in the San Francisco Bay Area, merging Lilly's 150 years of pharmaceutical expertise with Nvidia's dominance in AI computing infrastructure.
For decades, wet lab experiments and computational modeling operated in silos. Chemists synthesized molecules, and computers predicted properties. The feedback loop between discovery and validation stretched across months.
The Nvidia-Lilly lab eliminates that friction by creating a continuous learning system where robotic labs automatically generate experimental data, AI models learn from that data in real time, and the models immediately inform the next round of experiments, all without human intervention between cycles.
The architecture of the Nvidia-Lilly collaboration centers on what both companies call "closed-loop discovery", a scientist-in-the-loop framework where machines and humans work in parallel. Lilly's domain experts in biology, chemistry, and medicine will co-locate with Nvidia's AI researchers and engineers at the new facility, which is expected to open by the end of March 2026 in South San Francisco.
Instead of emailing files or scheduling meetings, researchers sit together, watching real-time data flow from wet labs (robotic systems physically synthesizing and testing molecules) into computational dry labs (Nvidia's BioNeMo platform training AI models on live experimental results). A chemist designs a protein target. AI models trained on Lilly's historical data generate candidate molecules. Robots synthesize those candidates. Experiments validate or reject them. AI models incorporate the new data. Hypotheses are refined. The cycle repeats, not weekly or monthly, but continuously.
"The focus will be on creating a laboratory where AI software (the dry lab) and robotic hardware (the wet lab) talk to each other 24/7, keeping humans in the loop without the need for constant reprogramming the robot or manually completing every step of the experiments," an Nvidia spokesperson told the technology press.
Lilly brings 150 years of pharmaceutical R&D expertise while Nvidia provides the BioNeMo platform and computational infrastructure to build next-generation foundation models for biology and chemistry.
The Hardware Advantage
The timing of this announcement is significant. Five days before Nvidia announced the Lilly partnership, the company unveiled Vera Rubin, its next-generation AI computing platform launched at CES 2026. Vera Rubin delivers five times more AI training power than Nvidia's prior-generation Blackwell GPUs while requiring only a quarter of the chips and costing one-seventh the token expense for inference.
Nvidia claims the Rubin platform can train a massive "mixture of experts" model (specialized AI agents) in the same timeframe as Blackwell while reducing the number of GPUs required. This efficiency translates directly to cost, training large biomedical foundation models on Lilly's datasets becomes materially cheaper with Rubin than with Blackwell.
The Nvidia-Lilly lab will be among the first to access Vera Rubin in volume. Nvidia began full production in Q1 2026, with cloud availability expected by the second half of 2026. Lilly's investment in the partnership essentially guarantees early access to this hardware, a competitive advantage worth tens of millions of dollars annually.
Both companies signaled intent to apply AI across Lilly's entire business including manufacturing, clinical development, and commercial operations. Nvidia will integrate three physical AI and robotics platforms, Omniverse (for digital twins), Isaac (for robot learning), and Jetson (for edge computing) into Lilly's manufacturing facilities.
Lilly manufactures GLP-1 receptor agonists (Mounjaro, Zepbound) at massive scale, facing constant constraints around production capacity and supply chain reliability. Digital twins created in Omniverse allow engineers to model, stress-test, and optimize manufacturing lines before making physical changes, reducing experimental waste and accelerating deployment of efficiency improvements.
Nvidia is simultaneously announcing partnerships with Thermo Fisher Scientific to build autonomous laboratory infrastructure and with Multiply Labs (a cell therapy manufacturing startup) to deploy robotic systems that could reduce cell therapy manufacturing costs by 70 percent per dose.
Competitive Context and Timeline
This partnership follows Lilly's announcement last October of a supercomputer built with Nvidia's Blackwell GPUs, the most powerful AI system ever deployed in pharmaceutical R&D, using more than 1,000 B300 GPUs. That system trained Lilly's foundation models for drug discovery and continues to power TuneLab, the company's open-access platform that gives biotech startups free access to Lilly-trained AI models.
Rather than simply investing in larger hardware, the companies are restructuring the laboratory environment itself. Traditional pharma R&D operates as a series of functional silos, chemists design molecules, biologists test them, and safety teams evaluate toxicity. The Nvidia-Lilly model collapses these silos into a single continuous system where all expertise converges in real time on the same problem.
Schrodinger also just announced integration of Lilly's TuneLab into LiveDesign, becoming the priority platform partner for external biotech using Lilly's AI models. Revvity announced a similar integration. These partnerships acknowledge a market reality that Lilly is winning the race to democratize access to pharma-grade AI capabilities.
The broader implications compound over time. If the Nvidia-Lilly lab successfully compresses drug discovery time, the impact extends far beyond R&D efficiency. Diseases with no approved treatments become addressable. Rare genetic disorders move from "economically infeasible" to viable. Cancer therapies can be personalized and manufactured faster than disease progression.
Neither company used this language in their official announcement, but it's implicit in every technical detail. Jensen Huang, Nvidia's CEO, called the partnership "a new blueprint for drug discovery, one where scientists can explore vast biological and chemical spaces in silico before a single molecule is made."
"Combining our volumes of data and scientific knowledge with Nvidia's computational power and model-building expertise could reinvent drug discovery as we know it," said David Ricks, Lilly's CEO.