American academics are sounding the alarm that staff reductions and anticipated budget cuts at the National Science Foundation are handicapping one of the Trump administration’s touted priorities: AI research and development.
For decades, the NSF has been a leading funder of AI innovation in academe, with a budget of more than $530 million for AI research in fiscal 2024 alone. Its grant programs include those advancing research, infrastructure, work-force development, and education around artificial intelligence.
Much of this work, AI experts point out, can’t be easily handed off to industry. Academics are often the ones conducting the time-intensive, basic research behind the products and services that companies ultimately create and bring to market. So federal support is critical, they say, to preserving the United States’ position as the global AI leader. It’s also critical for capitalizing on the current momentum that often eluded the field in the decades before ChatGPT.
“I’ve been in AI for 30 years” said Pascal Van Hentenryck, director of the AI hub at the Georgia Institute of Technology. “This momentum is very unique.”
Yet it is support that now feels uncertain. The NSF has lost subject-matter experts and some program officers, according to AI-grant recipients who spoke on background with The Chronicle. And while the approved government-spending bill earlier this month held research funding flat through September, further reductions in the NSF’s workforce are looming, and rumors abound that the agency could see a budget cut as deep as 66 percent — from $9 billion to about $3 or 4 billion — in fiscal 2026. (The president’s delayed budget proposal is now expected in April or May.)
Uncertainty alone complicates planning and delays decisions, sources said. It can thwart ambition. And it can drive away already limited talent in droves.
“Uncertainty is bad for the work force, it’s bad for policy, and it’s bad for the economy,” said Erin Mote, CEO of the nonprofit InnovateEDU. “It’s just really bad.”
Assessing the Damage
The Chronicle reached out to nearly four dozen principal investigators of AI-related grants that the NSF currently supports. The majority come from the National Artificial Intelligence Research Institutes and the ExpandAI program, the latter of which promotes minority-serving institutions’ participation in AI innovation.
The work these grants facilitate is expansive. Some recipients are driving research on trustworthy and explainable AI. Others are studying how to apply AI to areas like disease diagnoses, energy infrastructure, agriculture, and robotics. Still others are building up education and work-force development around AI, creating virtual teaching assistants, AI micro-credentials, and certificate programs.
The 10 principal investigators who responded shared varying accounts of what the last two months have been like. All requested anonymity, fearing possible retaliation against their grants.
Several said they were already feeling the fallout from the recent turmoil. Two ExpandAI grant recipients shared that they’d submitted “concept outlines” for the next stage of the program earlier this year, but have since heard that reviews are on hold. (Virtual office hours for the program appear indefinitely canceled as well.) Two leaders of separate AI Research Institutes said they’d lost at least one program officer — experts who can help grant recipients conceptualize and design research projects and offer constructive feedback throughout the grant period.
Most spoke to the stress of the unknown. One academic whose multi-year AI grant relies on support from both the NSF and the Education Department’s hollowed-out Institute of Education Sciences is anxiously awaiting word this summer on whether his grant will still be funded at the awarded amount. Another said they’d recently applied for grant renewal, but had been advised that such a renewal could come with a reduction in funding.
“We desperately need research to best understand how to use AI productively” and how to incorporate it into classrooms, one grantee wrote in an email. “It is really the worst possible time to be doing this.”
(The Chronicle reached out to more than a half dozen NSF program directors to help validate sources’ anecdotes. One responded via email, stating: “We have been advised by our union not to talk to the Press right now while various negotiations are taking place.”)
Sources said they were frustrated not only by the uncertainty surrounding their own work, but also by the broader implications for American competitiveness and the talent pipeline.
“It’s not like we’re far ahead of all of our competitors and are able to slow down. We’re maybe a couple of months ahead,” said the leader of one AI Research Institute. The recent release of two high-performing, comparatively low-cost AI models by the Chinese company DeepSeek, for example, sent shockwaves through the tech industry, calling into question the size of the United States’ lead in the AI race. Top economic officials in China have since promised to invest billions in emerging industries like AI.
The risk of “brain drain” is also a growing concern. Mote, at InnovateEDU, said she can “count on two hands, maybe more” the number of fired NSF employees — some of them AI experts — who’ve told her they would not return, despite several already receiving offers to come back. An AI Institute leader also told The Chronicle that they’d lost two postdoctoral students in the last two months to more stable industry positions. (Recruiting postdocs has been a persistent challenge; just 25 percent of more than 400 new Ph.D. recipients in AI/machine learning went into academe in 2023, according to the Computing Research Association’s 2023 Taulbee Survey.)
“This will have generational consequences,” Mote said. “We have to acknowledge that humans do the work here.”
Addressing the Dissonance
It’s not uncommon for an incoming administration to go through a “realignment” period, Mote said. But she added that the speed and lack of transparency is abnormal. And the dissonance between the administration’s stated embrace of AI and the sieges on agencies like the NSF, sources say, is especially bewildering.
President Trump has long positioned himself as a champion of AI research and work-force development. During his first administration, the National Artificial Intelligence Initiative Act of 2020 established the National Artificial Intelligence Initiative Office. The first cohort of the National AI Research Institutes kicked off that same year. Trump also proposed a more than 70-percent increase in the fiscal 2021 budget for AI research and development at the NSF, including funding for AI-related work-force initiatives at community colleges and minority-serving institutions.
“We considered the administration to be a helpful ally” when it came to investing in AI, recalled Margaret Martonosi, a professor of computer science at Princeton University. She led the NSF’s Computer and Information Science and Engineering directorate from 2020 to 2023.
Now, Trump’s second administration has its own stated mandate to “act decisively to retain leadership in AI.” It rolled back Biden-era requirements for companies developing and deploying AI, with officials including Vice President JD Vance underscoring the importance of not “paralyzing” AI development. The administration also recently compiled nearly 9,000 public comments — which, notably, do not appear to be publicly accessible, as they’ve been in the past — to inform an AI Action Plan that will “define the priority policy actions needed to sustain and enhance America’s AI dominance.”
So, what exactly is happening?
Sources have a few guesses. A new administration means a lot of new people. The anti-DEI agenda inherently runs counter to the mission underpinning many of the NSF’s grants, which is to encourage broader inclusion in emerging technologies. Observers also note that, as the loosening of restrictions implies, this administration is probably most interested in fostering industry-led AI innovation. (In January, Trump talked up a joint venture of up to $500 billion between OpenAI, Oracle, and SoftBank.)
The Chronicle asked Michael Kratsios and Lynne Parker — two leading officials within the federal Office of Science and Technology Policy — to explain how the administration’s current policy actions align with its strategic goals around AI and what alternative funding might be in the pipeline. Neither responded.
Pressing Forward
Sources weren’t all doom and gloom in their outlooks. Academics said they’re still eager to advance their projects and, if necessary, consider pursuing other sources of money.
Alternative funding streams do exist. Empire AI, a partnership between New York’s leading public and private universities, has announced 85 research projects through a philanthropic contribution from the Simons Foundation. Champlain College, a small, private institution in Vermont, plans to conduct research in partnership with Anthropic and is introducing a new AI-fellows program with help from the Argosy Foundation. Reed Hastings, the co-founder of Netflix — which leans heavily on AI to personalize user experience on the streaming platform — this week also announced a $50-million gift to erect the Hastings Initiative for AI and Humanity at Bowdoin College, in Maine.
Still, Shalin Jyotishi, founder and managing director of the Future of Work and Innovation Economy Initiative at the think tank New America, underscored that money from industry, foundations, and private philanthropy cannot fully replace government support.
“There is no substitute” for what the NSF provides, he said. Especially when it comes to foundational research, “industry will not pick up the bill. And philanthropy most certainly cannot.”
So what can researchers and academics do in this moment of instability? Jyotishi emphasized the need to make clear how projects that the NSF supports make a difference, and align with “the priorities of those currently in power.”
Just a few weeks ago, Martonosi did just that, posting on LinkedIn about how an NSF grant decades ago ultimately led to the creation of “a little startup called Google” — now a key industry player in AI development. She subsequently heard from a number of people who said they’d had no idea.
“Not everyone is in a position to connect those dots,” she said. “Those of us who are need to help.”
