Lila Sciences and NVIDIA warn health funding cuts could strand AI drug discovery behind rivals

Washington “pulled tens of billions from national health funding” as healthcare and biotech hit its ChatGPT moment, and Lila Sciences and NVIDIA executives say the timing is brutal. At Fortune Brainstorm Tech in Aspen this week, Geoffrey von Maltzahn, co-founder and CEO of Lila Sciences, described the outcome as a structural competitive disadvantage if the U.S. falls behind “at a corporate level, at a sovereign level.” In other words, this is not a normal budget squeeze. It is a race-season problem, right when AI and biology are converging into a new productivity tool for science.

Kimberly Powell, vice president of healthcare at NVIDIA, sharpened the stakes: “If we defund now, while the rest of the world leans in, which Europe is leaning in, which a lot of the Asian countries are leaning in, we will be left behind.” The core argument from both executives is that AI drug discovery is not just another software upgrade. They frame it as a change in how the scientific method can be executed, more like an always-on loop that can ask questions, gather context, observe, reason, and act.

To understand why this matters to decision-makers, it helps to see how the industry actually behaves. Drug discovery has historically been capital intensive, slow, and full of uncertainty. Even with breakthroughs, the calendar is punishing. That is part of why the funding and talent question is inseparable from AI competitiveness. According to the source, AI drug discovery is a $3.25 billion market today, growing at roughly 26% annually and projected to top $10 billion by 2031. When markets grow at that clip, the winners are usually the ones who can compound learning fast, keep expensive experiments running, and iterate quickly when results come back.

The executives also point to a broader investment pattern that underscores the urgency. Demis Hassabis’s Isomorphic Labs raised a $2.1 billion Series B earlier this year, and the source characterizes capital as flowing into AI drug discovery. But money alone does not solve the bottleneck if timelines are long and finish lines keep moving. In other words, even when private capital is willing, public funding can shape the ecosystem: shared infrastructure, foundational research, and the broader “conditions” that companies need to move from prototypes to regulated products.

Lila Sciences is an example of what “moving fast” looks like when you apply AI directly to scientific workflows. Von Maltzahn leads Lila, described as a Flagship Pioneering spinout. Lila has raised $550 million to build what it calls “scientific superintelligence,” running AI systems that execute the scientific method around the clock across materials, chemistry, and life sciences. The source highlights a concrete result: Lila’s agents identified catalysts for splitting water into hydrogen and oxygen that outperform the precious metals the industry currently relies on. Even more telling, von Maltzahn said a third of those suggestions initially made no sense to his Caltech-trained team. The ideas were still tested, and the “highest-performing catalysts” on record came out of that set.

That anecdote goes straight to the second-order risk behind the funding warning. If the U.S. backs off at the moment AI is accelerating hypothesis generation and experiment selection, it is not just losing one program. It is losing feedback loops. The source quotes von Maltzahn saying the current “Claude Code-esque moment” where a new intelligence gets injected into science and changes it forever “may not feel imminent,” but that it is “right around the corner.” If that is true, then public underinvestment now could mean fewer trained systems, fewer datasets, and fewer iterations by the time the world catches up.

Powell’s contribution shifts the conversation from lab speed to regulatory physics, because drug discovery is only half invention and half proof. She described NVIDIA’s role as foundational, investing in open-source biology foundation models, antibody design models, and multimodal models that companies like Lila adapt for specific scientific goals. She also warned about a different kind of fragmentation: “If it’s only closed models that exist, there’s not a lot of ability to create the conditions that all of these applications across life sciences can thrive in the age of AI.” That matters for boards and execs because model access affects whether the ecosystem can validate results, reproduce them, and scale.

The unresolved issue is FDA sign-off on molecules designed by AI when you cannot trace back the training data that produced them. Powell sees a path forward through digital twin models of biology precise enough that regulators will accept in silico evidence, but she said, “We’re just not there yet.” That sentence is doing a lot of work. It implies that today’s AI systems still need humanly interpretable or regulator-friendly evidentiary bridges, and building those bridges takes time and money. So if U.S. funding cuts arrive now, companies may be forced to stretch budgets across longer verification cycles while competitors build earlier regulatory and evidentiary workflows.

For executives watching from adjacent areas of healthcare, the signal is clear. AI drug discovery is moving from a futuristic concept to an operational advantage, with global players leaning in and private capital stacking up. If public funding “pulled tens of billions” at precisely the wrong time, the lag can compound: slower experimentation, fewer repeatable validations, and later regulatory readiness. Von Maltzahn and Powell’s warning is ultimately about who gets to set the pace of science when the scientific method itself becomes programmable.