NSF Is Handing One Organization $35M to Run America's Shared AI Research Backbone — Inside the NAIRR Operations Center Competition

Nearly every debate about American AI competitiveness eventually collides with the same bottleneck: the compute, data, and models that make frontier AI research possible are concentrated inside a handful of large technology companies and elite universities. A graduate student at a regional public university, a researcher at a minority-serving institution, or a small-team scientist working on AI for drug discovery or climate modeling often simply cannot get their hands on the resources their better-funded peers take for granted. The National AI Research Resource (NAIRR) is the federal answer to that problem — a shared national infrastructure meant to democratize access to advanced AI research capabilities. And with solicitation NSF 25-546, NSF is now selecting the single organization that will run it.

The NAIRR Operations Center (NAIRR-OC) solicitation is unusual among federal grants. NSF anticipates one award of up to $35 million over five years — a winner-take-all competition where the prize is not money for your own science but the mandate and resources to operate national infrastructure on behalf of the entire U.S. research community. That structure makes it one of the most consequential, and least understood, AI funding decisions NSF will make this cycle. Understanding what NAIRR is, why it is transitioning now, and what an operations center actually has to do explains both who can win this and why it matters far beyond the winner.

From pilot to permanent

The NAIRR did not appear from nowhere. The NAIRR Pilot launched in 2024 and has, by NSF's own accounting, already connected over 400 U.S. research teams with computing platforms, datasets, software, and models — spanning fields as varied as agriculture, drug discovery, cybersecurity, and education. The pilot proved demand: researchers who could never otherwise access serious AI infrastructure used it to do real work. The question the pilot left open was institutional. A pilot is stitched together from volunteered resources and temporary coordination. A permanent national resource needs an operator — an organization with the frameworks, staffing, and management structures to run it as sustainable infrastructure rather than a demonstration.

That is precisely the transition NSF-25-546 is designed to execute. NSF announced the operations-center effort as part of moving NAIRR from its pilot phase into a coordinated, durable national program aligned with the White House's AI Action Plan. The operations center is the connective institution that turns a proof-of-concept into a standing capability. Whoever wins it inherits both the pilot's momentum and the responsibility of not squandering it.

What the operations center actually has to do

The NAIRR-OC is not a research center and not a compute cluster. It is an operating institution, and its responsibilities read accordingly. NSF frames three core mandates:

• Organizational leadership — establishing the operational framework, organizational management, and success metrics for the NAIRR, in coordination with NSF and federal partners. This is governance and accountability infrastructure, not science.
• Building capabilities — developing stakeholder support through partner interfacing, deploying a unified web portal that gives researchers a single front door to computational, data, model, and training resources, integrating disparate resources, and running community outreach so the researchers who most need access actually find it.
• Pilot coordination and transition — working with existing NAIRR Pilot partners while planning the shift to independent NAIRR-OC operational functions, so the transition does not drop the 400-plus teams already relying on the system.

Read together, these mandates describe an organization that must be equal parts systems integrator, program manager, partnership broker, and community builder. The single most visible deliverable — a unified web portal that centralizes access — sounds mundane but is the crux of the democratization mission. If a researcher at an under-resourced institution cannot navigate to the compute they need, the entire premise of NAIRR fails. The operations center exists to make access real, not merely available.

Who can credibly win $35 million to run national infrastructure

The eligibility terms are restrictive in ways that signal who NSF expects to apply. Only certain organization types may submit; each organization may submit only one proposal; and any individual may serve as PI, co-PI, or senior personnel on at most one proposal. Anyone whose biographical sketch appears in the proposal counts as senior personnel — a rule that forces applicants to be deliberate about which heavyweight names they attach, because those people cannot be spread across competing bids.

More telling than the formal rules is the nature of the work. Running the NAIRR-OC demands an organization with existing capacity in large-scale research computing coordination, a track record of managing multi-institution partnerships, the staffing depth to operate a national service, and the neutrality to serve the entire research community rather than advantage its own campus. That profile points toward established research-computing consortia, major university systems with national-infrastructure experience, or nonprofit intermediaries that already operate shared scientific resources. A single lab, however brilliant, cannot credibly claim to run national infrastructure. This is an institutional competition, and the institution's operating maturity — not the elegance of any research idea — is what will be judged.

Why this matters even if you will never apply

Very few organizations can compete for a single $35 million infrastructure-operations award. So why should the broader grant-seeking community care? Because the NAIRR-OC determines who gets to do AI research in America over the next decade. If the operations center succeeds — if the portal works, the resources integrate, and outreach genuinely reaches under-resourced institutions — then thousands of researchers who are today locked out of serious AI work gain a viable on-ramp. That reshapes the competitive landscape for every downstream AI grant, from NSF's research institutes to agency SBIR programs, because it widens the pool of institutions capable of producing fundable AI research.

In other words, the NAIRR-OC is upstream of the AI funding you may actually chase. The organization that wins it becomes a gatekeeper and enabler for the entire ecosystem. For the broader context of NSF's AI investments and the FY2026 budget that funds them, see our Granted News coverage. And for institutions weighing whether they belong in this competition, the honest test is simple: can you credibly run a service the whole nation depends on? If the answer is not an immediate yes, the more productive path is to become one of the 400-plus teams NAIRR serves — and to build the record that might, in a future round, make you the operator.

The details, including the complete eligibility list and submission timeline, live in solicitation NSF 25-546; questions go to nairr-oc@nsf.gov. Whoever wins will spend the next five years deciding how open American AI research really becomes.