On August 4, 2026, the National Science Foundation opened NSF 26-513, the State and Regional Artificial Intelligence Infrastructure Hubs solicitation — a $100 million cooperative-agreement programme intended to put frontier AI compute within reach of researchers outside the small number of institutions that already have it. Full proposals are due November 4, 2026, with the same first-Wednesday-in-November deadline repeating annually.
The programme is structured as a public-private consortium model: NSF funds the coordination and workforce layer, while consortium partners — states, industry, philanthropy — are expected to fund the compute itself. That split is the single most important thing for a prospective applicant to understand before starting to build a consortium, and it is where this guide starts.
What NSF just opened
NSF plans to make up to 10 awards, each worth $4 million to $12 million over five years, for a total of roughly $100 million in this initial round. Awards are structured as cooperative agreements rather than standard grants, meaning NSF retains substantial involvement in how each hub is run, not just in how it’s funded. Only one award will be made per state or multi-state region, which is the constraint that will do the most to shape who partners with whom.
NSF frames the hubs as connecting to two other initiatives already in motion: the National AI Research Resource (NAIRR) pilot and the White House’s Genesis Mission. TechAccess: AI-Ready America (NSF 26-508) and the Campus Research Computing Consortium are the other adjacent programmes NSF points to as related infrastructure investments.
Who can lead a proposal
Eligibility is broader than a typical NSF research solicitation, deliberately so — the stated goal is to reach institutions that don’t already have frontier compute access.
- Institutions of higher education — both two-year and four-year, including community colleges — that are accredited in the United States.
- Non-profit, non-academic organizations based in the U.S., which NSF’s solicitation lists as including museums, observatories, independent research laboratories, and professional societies.
Two restrictions matter for consortium planning:
- Each organization may participate in at most one proposal per deadline, whether as lead or partner.
- Each PI or co-PI may appear on at most one proposal per deadline, across all roles.
Because only one award can go to a given state or multi-state region, institutions that might otherwise compete against each other have a direct incentive to consolidate into a single regional consortium rather than submit rival proposals that are guaranteed to leave at least one unfunded.
What a consortium must include
A proposal must be built around a consortium, not a single institution. NSF requires that the consortium include at least one institution of higher education, and strongly encourages — without formally requiring — that it also include private industry, philanthropic organizations, and state or local government. NSF also strongly recommends that each hub establish an external advisory board.
What NSF funds versus what the consortium must fund
This is the distinction most likely to catch a first-time applicant off guard: NSF money in this programme does not pay for the compute itself.
| NSF funding covers | Consortium partners must fund |
|---|---|
| Consortium coordination | Acquisition of new or expanded computing, data, and software infrastructure |
| AI infrastructure professionals — technical staff who support researchers directly | Ongoing operations and maintenance of that infrastructure |
| Faculty training and instructional material development | All funding for new or expanded computing, data, and AI resources across the full five-year period |
| Workforce development for “AI for science” facilitators | — |
In practice, this means the NSF award functions as seed funding for the human and coordination layer of a hub, while the hardware and infrastructure spend has to come from the state, industry, or philanthropic partners in the consortium. A proposal that hasn’t lined up a real infrastructure funding commitment from a non-NSF partner before submission is unlikely to be competitive, since that commitment is what the NSF funds are designed to sit on top of.
Award structure and key dates at a glance
| Item | Detail |
|---|---|
| Solicitation number | NSF 26-513 |
| Total programme funding | ~$100 million |
| Number of awards (this round) | Up to 10 |
| Award size | $4M–$12M per hub, over 5 years |
| Award instrument | Cooperative agreement |
| Geographic limit | One award per state or multi-state region |
| Full proposal deadline | November 4, 2026, then the first Wednesday in November annually |
| Lead eligibility | U.S.-accredited 2- and 4-year institutions of higher education; U.S. non-profit non-academic organizations |
The industry and philanthropic partners already involved
Alongside the solicitation, NSF named a set of industry and philanthropic partners supporting the initiative: NVIDIA, AMD, Intel, Dell Technologies, Hangar, and the Secunda Innovation Fund. NSF’s announcement did not specify which, if any, of these partners are pre-committed to a specific regional hub; consortia are expected to identify and secure their own regional industry and philanthropic collaborators as part of proposal development, and named national partners appearing in NSF’s own announcement do not substitute for that regional partnership-building.
Brian Stone, performing the duties of the NSF Director, said the programme reflects that “artificial intelligence is transforming how we conduct research, accelerate scientific discovery and address complex challenges across disciplines.” White House Office of Science and Technology Policy Director Michael Kratsios framed it as a matter of tooling equity, saying “American scientists deserve the world’s best tools to enable their most ambitious work.”
Why this matters for institutions outside the AI frontier
The programme’s design — broad lead eligibility including community colleges, a one-award-per-region cap, and an explicit non-NSF funding requirement for the compute itself — signals that NSF is trying to solve a distribution problem rather than simply add more capacity at institutions that already have GPU clusters. For research administrators at institutions that have not historically competed for large infrastructure awards, the most consequential planning task is not the NSF proposal narrative itself but identifying and formalizing the regional partnership — state government, industry, or philanthropic — that will fund the actual infrastructure NSF’s award will not cover.
Because only one hub can be funded per state or multi-state region, institutions considering separate proposals within the same region should treat early coordination as the higher-priority task, given that competing proposals from the same region cannot both be funded regardless of individual merit.
Frequently asked questions
Does the NSF award pay for the GPUs or compute infrastructure itself?
No. Per the solicitation, NSF funding covers consortium coordination, technical support staff, faculty training, and workforce development. All funding for acquiring, operating, and maintaining new or expanded computing, data, and AI infrastructure must come from the consortium’s own partners over the five-year award period.
Can a single institution apply without forming a consortium?
No. The solicitation requires a consortium built around at least one institution of higher education, and strongly encourages including private industry, philanthropy, and state or local government as partners.
How many hubs can be funded in one state?
Only one award will be made per state or multi-state region in this round, which is why NSF’s solicitation is likely to push would-be competitors within the same state toward a single joint proposal.
When is the deadline?
Full proposals are due November 4, 2026. NSF’s solicitation states the deadline recurs on the first Wednesday in November in subsequent years.
Is this the same programme as NSF’s Regional Innovation Engines?
No. NSF Regional Innovation Engines (NSF Engines) is a separate, earlier NSF programme aimed at regional innovation ecosystems more broadly. NSF 26-513 is specifically focused on AI compute infrastructure access.
