NVIDIA Bets Big on AI-Driven Drug Discovery, Physical AI, and a $1 Billion Eli Lilly Partnership

By Allison Proffitt 

January 12, 2026 | NVIDIA is making a massive push into healthcare and life sciences, announcing a landmark five-year, $1 billion strategic partnership with pharmaceutical giant Eli Lilly alongside major expansions of its biological AI platform today at the J.P. Morgan Healthcare Conference. 

Foundational to this push into life sciences is an expansion of BioNeMo, NVIDIA’s large language model for biology announced in the fall of 2022. Kimberly Powell, NVIDIA's VP of Healthcare and Life Sciences, described the update as a “major expansion … to a full open development platform that enables lab-in-a-loop scientific workflows for biology and drug discovery.”  

Facetime Facility  

The $1 billion Lilly partnership will bring this significant digital update into physical use with a new AI co-innovation lab in the San Francisco Bay Area where scientists from both Lilly and NVIDIA will work side-by-side on accelerated drug discovery, clinical development optimization, and advanced manufacturing applications.  

“This will be a site that our scientists will join together and physically co-locate,” Powell said. “A lot of that cross pollination can happen in real time… and we’re deeply excited. We’ve fully allocated dedicated incremental resources, in fact, to this partnership.” She says the facility is expected to open by late March. 

The collaboration is in addition to an October announcement that Lilly and NVIDIA were collaborating to build the pharmaceutical industry’s most powerful supercomputer. The new billion-dollar commitment will provide additional computational resources beyond that “AI factory” already being set up, according to Powell. 

“By combining Lilly's deep domain expertise in drug discovery and NVIDIA's expertise in AI and accelerated computing, we are building the future of how medicines will be designed and developed,” she added. 

Open Models Fuel Healthcare AI Boom 

NVIDIA is the world's largest contributor of open-source AI on HuggingFace, Powell said in the briefing, releasing over 650 open models and 250 open datasets across biology, chemistry, medical imaging, and other domains in 2025 alone. 

“We are tripling down and building open models, open datasets, and open-source tools to facilitate the exponential progress in the domain-specific field of life sciences,” Powell said. “Open models are the backbone of modern innovation.”  

The strategy appears to be paying off. Powell noted that healthcare is deploying AI at nearly three times the pace of the broader U.S. economy, with 80% of startups building on open models. Companies like Abridge, which uses AI to cut clinical documentation work by over 30% and now serves 200-plus healthcare systems, are running on NVIDIA's NeMoTron platform. 

Autonomous Labs Take Shape 

But NVIDIA isn’t limiting itself to models and digital advancements. With Thermo Fisher Scientific and robotics firm Multiply Labs, NVIDIA is creating increasingly autonomous physical laboratory environments. 

Thermo Fisher, the world leader in lab equipment and services, is integrating NVIDIA AI platforms directly into instruments and workflows. Using the compact DGX Spark AI system as a “benchtop supercomputer,” the companies have built AI agents that can help design experiments, control instruments via APIs, monitor quality in real-time, and analyze results. 

In one of Powell’s examples, a flow cytometer equipped with an AI agent can detect when a cell sample has clogged the instrument and automatically unclog it—a closed-loop quality control system requiring no human intervention. 

Meanwhile, Multiply Labs is using NVIDIA's Omniverse and Isaac platforms to create digital twins of entire laboratories, training robots on thousands of precision tasks in simulation before deploying them in actual facilities. The results in cell therapy manufacturing have been dramatic: a 70% reduction in cost per dose and a 100-fold increase in throughput per square foot. 

Biology's "Transformer Moment" 

Powell predicted 2026 would be biology's "transformer moment.” NVIDIA’s expanded BioNeMo platform will include new models for RNA structure prediction, molecular synthesis, toxicity prediction, and more.  

The “lab-in-the-loop” workflows—continuous cycles where experimental data feed directly back into AI model training—is “not just operational improvement,” Powell said. “It’s a foundational enabler of more intelligent biological AI.”  

The system generates new hypotheses for laboratory testing and AI agents coordinate the work across digital and physical labs including automatically creating instrument protocols.  

Companies are already building on this infrastructure. Basecamp Research is launching EDEN, a gene therapy design platform trained on 10 trillion biological tokens. TetraScience has adopted multiple BioNeMo models to help biotech companies analyze scientific data. And startups like Edison Scientific have created AI systems capable of deploying 200 agents to synthesize discoveries from 1,500 scientific papers—work that would take humans months—in less than a day. 

The Bigger Picture 

The announcements reflect NVIDIA's strategy of providing end-to-end AI infrastructure—from open models and training frameworks to edge computing hardware and simulation platforms—specifically tailored for life sciences applications. 

Whether this approach can deliver on the promise of dramatically accelerating drug discovery and reducing costs remains to be seen. But with healthcare facing projected shortages of tens of millions of professionals by 2030, according to World Health Organization estimates, and pharmaceutical R&D costs continuing to climb, NVIDIA is ensuring their position.