NSF’s AI agenda: Aiding Genesis Mission, breaking down adoption barriers
The National Science Foundation is working to mitigate roadblocks to AI adoption — such as a lack of GPUs and cost constraints — as it aims to play a supporting role in the ongoing national effort, according to a top agency official.
The NSF has long been a cornerstone of national technology interests, from democratizing biotechnological research infrastructure to creating open multimodal AI models for the scientific community and investing in AI-ready test beds. With the Energy Department at the helm of the AI-focused Genesis Mission, NSF is honing its strategy to champion the efforts.
“The short answer is it’s a work in progress,” Ellen Zegura, NSF’s acting assistant director of computer and information science and engineering, said of the relationship between her agency and DOE during a public information gathering session last week. “We see both complementary and collaborative elements to it.”
The Genesis Mission has several broad goals, such as creating a national AI platform and improving the nation’s productivity tied to the research-and-development budget. Zegura said NSF wants to find ways to bolster the mission as well as identify areas of opportunity that aren’t necessarily covered by the executive order to launch the Genesis initiative.
Building up the nation’s AI workforce and contributing to skills development, for example, is one area that NSF continues to have its eye on. The agency published a roadmap last year, aiming to outline key skills and investments needed to facilitate needed STEM skills. NSF sought input from organizations and individuals across sectors on the identified pathways and recommendations.
Lawmakers have also tried to expand the agency’s workforce and training remit. A bipartisan trio of House members re-upped legislation in September that would authorize NSF to grant AI-related scholarships and fellowship opportunities.
Despite its ambitions, NSF faces the same barriers to AI adoption that many agencies and private-sector organizations are running up against: constraints on the chips that power AI and runaway costs associated with the technology.
The demand for AI chips continues to soar at an unprecedented pace. Nvidia, a leading supplier of the beefed-up processors, has seen its revenues balloon and its supply stretched thin.
“The clouds are sold out and our GPU installed base, both new and previous generations, including Blackwell, Hopper and Ampere, is fully utilized,” Colette Kress, EVP and CFO at Nvidia, said during the company’s Q3 2026 earnings call in November.
Reining in costs is also top of mind for most organizations trying to move adoption forward. Analyst firm Gartner estimates that there are 10 hidden costs for each AI tool organizations buy, as well as transition costs associated with training and change management.
NSF is working to find a resource-conscious middle ground.
“There will be capabilities and capacities that are at the very highest end that NSF will be priced out of,” Zegura said. “It’s really important that there be broad ability to access that next level down because that’s where we provide more opportunity for newer or disruptive ideas to come about.”
The agency plans to facilitate access for a wide set of researchers in the “one level down” range, Zegura added. If the NSF is priced further out, the potential positive impact could dwindle.
“It doesn’t have to be at the uppermost level, but you have to be in shooting distance,” Zegura said. “Otherwise you don’t have a trajectory, a path from the testing of an idea to that of a big playing field.”
Reorganization, refocus
Like other agencies, NSF has gone through palpable change in the last year.
The agency lost more than 500 workers in 2025 and has already shed a handful this year, according to the Office of Personnel Management’s newly launched federal workforce data site. Out of those workers, 30 were classified as IT managers.
The required reconfiguration at NSF has resulted in a bundled office of advanced cyberinfrastructure, research infrastructure and other large projects, including the AI institutes, Zegura said.
“We’ve made changes that bring the AI institute program officers closer to the [Office of Advanced Cyberinfrastructure] program officers,” Zegura said. The evolved structure could enable experiments around alternative AI architectures that may not have occurred when the groups were further apart, she added.
The AI institutes have been largely viewed as a success and a template for forthcoming initiatives, according to Zegura. The AI institutes date back to 2019 but have evolved from their earliest iterations.
“It’s really been extremely successful, so we’re inclined to continue doing it,” Zegura said. “It’s produced some exciting research, some really interesting education efforts and outreach — sort of hits on all the things we love to see.”
The AI institutes are powered by partnerships, which “is not without challenges,” according to Zegura. The partner awards are often lengthy commitments, which can become prickly due to turnover or the rate of innovation. The agency is learning from past mistakes, Zegura said, and reworking partnership structures.
Still, the AI institutes have been able to fill critical gaps, Zegura said.
“In the beginning, everything was a gap,” she said. “Now, we’re wanting to be as strategic as possible about where we make the additional investments and really interested in getting input … from communities.”
