The journey from molecule to medicine requires years of laboratory work. "There are more medicinal molecules out there than stars in the universe,” said Eli Lilly and Company's Executive Vice President and Chief Information and Digital Officer Diogo Rau from The Big Interview stage on December 4. “But finding one that works is about as rare as finding a single star in a galaxy, like finding one star in the Milky Way.” Researchers test molecules against hypotheses or known biological targets (like blocking a pain receptor) and see what works. They do that millions of times to find the single molecule that will lead to a health outcome.
But thanks to AI, we may be able to shrink that timeline—and end illness altogether.
AI as a Lab Partner
Equipment sterilization and worker safety notwithstanding, the trial-and-error process of developing new pharmaceutical products has been roughly unchanged for the better part of the last century. But according to Rau, AI is not only speeding up the timeline of discovery, but giving researchers new ways to approach it. Adding generative AI into the exploratory process of drug research has helped surface novel solutions. In fact, the first time Rau’s team used generative design on molecules, they produced some results that were promising, but unlike anything else the lab had ever come across. Intrigued, the team ran these experiments by Lilly staff chemists, and the response surprised Rau: “They said, 'That's interesting. We hadn't thought about doing it that way.' So the AI actually inspired the chemists to go down a different path that they hadn't been exploring.”
This series of events speaks to a broader shift happening throughout the professional world—human-machine collaboration. Just as knowledge workers in other industries have had to learn how to incorporate AI into their workflow to boost efficiency and drive faster results, the sciences are no different. "We always talk about how we're training our machines," Rau said. “But there's a lot that they're going to teach us on the human side that we're going to go explore… not because we trained them, [but] because they're actually teaching us.”
An Efficiency Booster
In the early years of artificial intelligence research, one of the oft-stated benefits of AI was faster and more accurate results in medical labs and scientific research. While Lilly has been on the forefront of applying computing power to pharmaceutical challenges since becoming the first commercial purchaser of a Cray-2 supercomputer in 1989, a strategic partnership with a market leader in AI chips will enable them to unlock unprecedented gains in research volume. Coming online in January 2026, Lilly's new technology stack will use purpose-built models trained on internal data to launch scientific AI agents (digital assistants supporting researchers), push forward drug discovery research, and advance manufacturing operations via virtual modeling to fine-tune supply chains.
Critically, this custom-built LLM will learn from past and future molecular experiments. According to Rau, "There are more medicinal molecules out there than stars in the universe, but finding one that works is about as rare as finding a single star in a galaxy, like finding one star in the Milky Way." With this much computing power, the dream of finding cures for as-yet incurable diseases is tangible. Adding AI to the equation speeds up that timeline significantly. Rau said that on the first day their supercomputer is online, it will be 7 million times as powerful as the Cray-2 supercomputer.
That kind of hyper acceleration forever changes the time horizon for pharmaceutical companies and their output. Due to the Hatch-Waxman Act of 1984, pharmaceutical patents expire 20 years after their filing date, allowing generic drugs to hit the market. That puts drugmakers in a constant cycle of discovery, testing, and go-to-market. "We've got to be focused on what we're going to discover," Rau said.
Installing Guardrails to Accelerate Growth
While this kind of accelerated pace may seem anathema to the slow progress of safe science, Rau notes that Lilly has built governance best practices into their systems at every level, building a platform-agnostic system with the capacity to utilize up to 43 large language models (LLMs), with guardrails to protect against issues like exposure of personally identifiable information. While today, developers build out new products and systems, according to Rau, “The real vision is [that] anybody in any role… can just go and talk to a large language model and have the guardrails and make everything work.”
A Moon Landing Moment
The pace and fervor of this innovation can have tangible results. Even without AI, the scientific community has eradicated smallpox globally and turned once-fatal diseases into manageable conditions. With the support of artificial intelligence working in collaboration with lab researchers and civil society, Rau envisions a dramatic future. In ten years, we expect to see new medicines that have been shaped by AI. "But," he said, "by the end of the century we essentially could be disease free as we know it." That is the true promise of AI.
Learn more about Eli Lilly and Company at www.lilly.com
