As AI bolsters the drug discovery pipeline, the boundaries of pharmaceuticals and technology are becoming increasingly blurred. Pharmaceutical giant Eli Lilly's unveiling earlier this year of its supercomputer LillyPod, developed in partnership with NVIDIA, epitomizes this evolution.
The Indianapolis-based AI factory LillyPod opened in February and is built with 1,016 of NVIDIA's high-powered Blackwell Ultra GPUs. It functions as a computational dry lab that works in tandem with Lilly's San Francisco-based agentic wet lab (where tangible drug development is happening) to speed up research and development.
"NVIDIA isn't entering drug discovery, and we're not entering GPU design," said Diogo Rau, Lilly's executive vice president and chief information and digital officer. Even so, this collaboration marks the maturing of a hybrid industry now perpetually entangled.
Other partnerships, like those between the likes of Johnson & Johnson and Microsoft Azure, or Pfizer and Google Cloud, have been brewing for years. But this new phase moves beyond cloud infrastructure into integrated, purpose-built AI hardware and co-innovation.
"The perception that pharma is anti-tech is dated," said Erwin Estigarribia, CEO of Headlamp Health, whose intelligence platform Lumos AI augments clinical trials. "AI is simply the next layer, and it can finally do something with the decades of disenfranchised, disconnected research sitting across internal repositories, archived notebooks, and failed programs."
The first wins from these NVIDIA-class partnerships, said Estigarribia, will come from re-analyzing decades of legacy data in Lilly's possession, finding answers that were invisible the first time through. Only then will AI be fully trusted on novel prospective workflows.
Beyond the algorithm
As computing power reshapes the pharmaceutical industry, it simultaneously forces a reckoning with the structural and ethical guardrails needed to govern such a powerful force.
One concern is algorithmic bias, something Rau says Lilly remains keen on. "We don't claim any model is unbiased," he said. "We build systems with testing, auditability, transparency and human oversight so issues can be identified and corrected over time."
As for what this looks like in practice, some vendor tools require third-party bias audits. For others, Lilly sets explicit documentation requirements on mitigation strategies.
Medical AI is only as fair as the data it learns from, said Ardy Arianpour, CEO and founder of health data repository platform SEQSTER. "That data has historically excluded women, people of color, low-income populations and rural communities," he said.
That's why, according to William Soliman, who heads up the Accreditation Council for Medical Affairs, strong governance frameworks are essential. "This includes robust data quality standards, bias testing across diverse patient populations, explainability requirements for high-impact clinical decisions, continuous post-deployment monitoring, and clear accountability when errors occur."
According to Estigarribia, the combination of accelerating innovation and chronically constrained regulation agency resources makes it structurally impossible to stay ahead of risk. This, he said, is why the industry has to self-regulate. "I know this can sound like letting the fox guard the henhouse," he said, but we're seeing foundations arise.
Rau said Lilly aligns all processes with the U.S. National Institute of Standards and Technology's AI Risk Management Framework and keeps people in the loop at high-stakes decision points.
In the U.S., the Food and Drug Administration's Digital Health Center of Excellence and its plans for AI-enabled medical devices is a gradual start. The European Union's Artificial Intelligence Act, first enforced in August 2024, is considered the first binding legal framework designed to ensure AI-enabled medical devices are fair, transparent and worthy of clinical trust. "AI‑enabled medical devices regulated under the MDR [medical device regulation] and IVDR [in-vitro diagnostic regulation] are automatically treated as high risk, triggering hard obligations rather than ethical encouragement," said Linzi Penman, partner and UK Tech Sector lead at law firm DLA Piper.
The U.K., meanwhile, has opted for a different route. Rather than a single AI statute, it has empowered regulators to operationalize shared principles via the AI Regulation White Paper of 2023. "Sector regulators are expected to apply safety, transparency, fairness, accountability and contestability within their domains," said Penman.
The Information Commissioner's Office in the U.K. is clear that organizations must ensure AI systems do not learn, repeat, or amplify unfair biases, she added.
At the root of much of this movement is the Organisation for Economic Co-operation and Development, a global member group that updated its AI principles in 2024.
The success of partnerships like LillyPod depends not just on compute power, but on the oversight needed to ensure pharmaceutical AI serves the real driver of innovation: patients. "Patients generate the data that trains the models, funds the trials, and ultimately validates the science," said Arianpour. "They are the origin point of everything."
The new drug maker
As these industries continue to converge in game-changing ways, the workforce that props up the pharmaceutical industry is poised to change with it.
Rau said the new paradigm requires people who understand both the science and tech of it all. "I call it being bilingual," he said. "Computational biologists who can build models and interpret wet lab results, AI-native chemists who understand both gradient descent and synthesis routes, data scientists who can work alongside clinicians and know what a p-value actually means in a trial context." The people who bridge that gap are in enormous demand across the sector, he said.
"The field is no longer organized around disciplines," said Estigarribia. "It's organized around the ability to translate across them."
Consequently, as delineations blur between expertises and between industries, the future of pharmaceuticals positions the lab and the AI factory as inextricably linked, if not one and the same.
