Ai and Biotech are about to kick ass together.

Ai has the ability to greatly speed up the new drug development process. BAM!

Nvidia and Eli Lilly are launching an AI drug discovery lab

A multiyear co-innovation lab ties Eli Lilly’s drug pipeline to Nvidia’s next hardware cycle, betting faster discovery comes from tighter loops

By Shannon Carroll, Quartz Media

Updated 18 hours ago

The doctor will see you now. It’s an AI agent with a visitor badge — and an eight-figure monthly burn.

Nvidia and Eli Lilly announced Monday that they’re launching a co-innovation AI lab aimed at speeding up drug discovery — a five-year commitment of up to $1 billion, built around accelerated, closed-loop discovery that’s meant to “industrialize” AI models and accelerate clinical development. The companies will co-locate at a new Bay Area site so teams can work together in real time, with the opening targeted for the end of March.

Kimberly Powell, Nvidia’s vice president of healthcare, linked the scaling plan to Nvidia’s future hardware, saying Lilly’s newly deployed AI factory is expected to grow into a hybrid cloud environment powered by future Nvidia Vera Rubin systems, alongside Nvidia’s DGX cloud capacity. For Nvidia, that’s the kind of detail that turns a partnership into something sturdier than a headline: a multiyear plan that’s already looking past the current generation of machines.

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

Her description of what the lab will do leans hard on data generation — the unglamorous ingredient that determines whether “AI drug discovery” behaves like a discipline or a demo. She said a major focus will be on producing “amazing training data” through large-scale lab work, creating “ground truth data in the lab” to train biology foundation models with multimodal data, then tightening the loop between hypotheses and discovery. And that’s the premise behind the whole setup: better experiments create better data; better data creates better models; better models make the next experiments more targeted — a feedback loop engineered for throughput.

In a press release, Lilly’s chief information and digital officer, Diogo Rau, described the lab as a shift in how discovery gets done. “We see this as a catalyst for the capabilities that will define the next era of drug discovery,” he said. “By working with Nvidia, we’re uniting massive compute, specialized talent and the ability to shape data at immense scale.” He added: “We’re moving toward a future where discovery is driven by rapid experimentation and increasingly customized models.”

Powell’s version of the same bet leans hard on data creation — “ground truth data in the lab” produced through large-scale lab work — because biology models only get as good as the inputs you can defend. The scope goes beyond early R&D. Powell said the companies will explore applying accelerated computing and advanced AI across Lilly’s business, from manufacturing to commercial operations. That’s a wide mandate — and a familiar one for Nvidia, which tends to start with the sexiest workflow and then angle for the rest of the enterprise stack once the infrastructure is in place.

“We’re going to use all of the amazing work with our open models in Nemotron to unearth all of this clinical understanding, so clinical development can also be enhanced and streamlined as much as humanly possible,” Powell said.

The lab is an expansion on top of the infrastructure Lilly has already been building with Nvidia. In late October, Lilly announced it was building what it called the industry’s most powerful pharmaceutical-owned AI supercomputer using Nvidia’s DGX SuperPOD platform to power an “AI factory” for drug discovery and development.

AI companies have been racing to turn “health” into an actual product category instead of a vague promise, with OpenAI rolling out ChatGPT Health and Anthropic launching Claude for Healthcare within days of each other. That consumer-and-clinician push has the same underlying logic as Nvidia’s: Whoever owns the workflow owns the habit, and whoever owns the habit gets to sell the infrastructure.

Nvidia’s broader point at the JPMorgan healthcare conference, which starts Monday, is that healthcare is ready for the same full-stack takeover that it’s already pulled off in other industries. The company is also rolling out a Thermo Fisher collaboration aimed at wiring Nvidia’s AI platforms into scientific instruments and lab workflows, a Multiply Labs partnership that leans on robotics and digital twins for biomanufacturing, and an expanded BioNeMo platform that Nvidia says will ship more open models, tools, and training recipes for end-to-end biology and drug discovery work.

For Lilly, a five-year, billion-dollar commitment puts real pressure on results that pharma investors actually care about: shorter discovery cycles, more viable candidates entering clinical trials, fewer dead ends that soak up years and capital before quietly disappearing. AI has promised all of that before. This lab is Lilly saying it’s willing to be judged on whether the promise shows up in the pipeline, not just in demos.

That also sharpens the risk. Drug discovery doesn’t fail because ideas are scarce; it fails because biology is stubborn, regulation is slow, and validation is unforgiving. Models trained on “ground truth” still have to survive replication, scale, and FDA scrutiny. Automation still has to earn the trust of scientists whose careers are built on skepticism. If this lab works, it could become a template for how pharma actually builds medicines. If it doesn’t, it could become an expensive reminder that compute can’t negotiate with biology.

📬 Sign up for the Daily Brief