NTIA Is Spending the Last $53 Million of a $1.5 Billion Program on AI-Native RAN. Applications Close September 9.

A $1.5 billion program is ending with $53 million and a change of subject.

That is the honest framing of the fourth and almost certainly final Notice of Funding Opportunity under the Public Wireless Supply Chain Innovation Fund, which NTIA opened on July 14, 2026 and closes on September 9, 2026. The NOFO is titled "Solutions for AI-Native RAN." It makes up to $53 million available, caps any single project at $25 million, requires a minimum 50 percent cost share, and promises awards on a rolling basis beginning in the fall rather than in a single post-deadline batch.

Every one of those parameters is doing work. Read together, they describe a program that has been rescinded down to a fraction of its original size, redirected away from the technology it was created to fund, and handed to a very small number of applicants who can write a check nearly as large as the one they are asking for.

How $1.5 billion became $53 million

The Innovation Fund was created by the CHIPS and Science Act with $1.5 billion and a specific mission: break the global radio access network equipment oligopoly by underwriting open, interoperable, standards-based alternatives — open RAN — and strengthening the competitiveness of U.S. suppliers against Huawei, ZTE, and the incumbent European vendors.

It spent real money on that mission. Rounds 1 and 2 committed more than $550 million across 35 awards, with Round 1 alone distributing $140.4 million across 17 grants. The money went to testing and evaluation labs, radio unit commercialization, and hardware R&D — the unglamorous plumbing of an interoperable network stack.

Then two things happened in quick succession. Congress rescinded $850 million of the fund through the One Big Beautiful Bill Act. And in late February 2026, NTIA Administrator Arielle Roth — the Assistant Secretary of Commerce for Communications and Information — used a February 25 address at the Media Institute's Communications Forum to signal that the remaining balance would be redirected toward AI-native wireless networks rather than continued open RAN hardware work. NTIA followed with a listening session, noticed in the Federal Register on March 11, 2026, and then confirmed the consequence that matters most to anyone who spent months on a proposal: NTIA will not issue awards under the Innovation Fund's third NOFO.

If you are one of the organizations that responded to Round 3, that is the single most important fact in this article. Work product from a canceled round is not wasted, but it is not an application either. The technical volumes, the teaming agreements, the letters of commitment — those are inputs to a Round 4 submission that has a different scope, a different evaluation frame, and three weeks left on the clock.

What "AI-native RAN" actually means in the NOFO

The scope shift is not cosmetic. Rounds 1 and 2 asked whether you could build interoperable radio hardware and prove it worked in a lab. Round 4 asks whether artificial intelligence embedded in the RAN can make a network cheaper to run and more valuable to operate — and whether you can prove it outside a lab.

The published priorities cluster around four requirements:

• Real-world demonstrations that generate data confirming business case projections. Not simulations. Not testbed results. Field data that validates a financial model.
• Compliance with 3GPP and O-RAN specifications. The open, interoperable premise of the original fund survives; AI is being layered onto it, not substituted for it.
• Demonstrated operational cost reduction and new revenue streams. The NOFO explicitly asks how the architecture changes operator economics, which is a different question than whether the technology works.
• Cybersecurity integrated throughout design and operation, not bolted on at the end.

The strategic frame NTIA has put around this is export policy as much as technology policy. The announcement ties the funding to "promoting a trusted, exportable U.S. AI technology stack where American innovation can compete globally," and Roth has framed the goal as enabling American innovators to showcase commercial use cases of 6G that demonstrate to investors and allied governments that an AI-native approach beats what Chinese vendors are selling.

That framing is a scoring signal. A proposal that reads as a domestic R&D project will land differently than one that reads as a demonstration designed to be shown to a foreign carrier procurement committee.

The cost-share arithmetic decides who can actually apply

The 50 percent minimum cost share is the eligibility filter that does not appear in the eligibility section.

Run the numbers. A $25 million award — the ceiling — requires at least $25 million in matching contribution, for a total project value of $50 million or more. Even a modest $8 million request obligates you to $8 million of your own money or committed third-party funds. Across the full $53 million, NTIA is buying at least $106 million of total project activity, and probably more, because competitive applicants routinely exceed the floor to signal commitment.

That arithmetic tells you the realistic applicant pool: equipment vendors with balance sheets, carriers, chipmakers, well-capitalized systems integrators, and consortia in which a large commercial partner is carrying the match for a smaller technical lead. A university lab or a seed-stage startup cannot meet this alone. It can, however, be the technical core of a team whose prime has the capital — and in a program explicitly about commercialization and export, that teaming structure is not a workaround. It is the intended shape.

It also tells you how few awards there will be. At the $25 million ceiling, $53 million supports two projects. At $10 million average, it supports five. The realistic outcome is somewhere between two and six awards, which makes this one of the most concentrated federal competitions running this fall. There is no consolation tier.

Rolling awards change the submission calculus

Most federal competitions are batch-scored: everything closes, everything gets reviewed against everything else, and awards land together. NTIA has said it expects to make awards on a rolling basis beginning in the fall.

Rolling awards in a fixed-pool program create a first-mover dynamic. Money obligated in October is not available in December. Whether NTIA formally reserves capacity across the review window or not, an application that arrives early, scores well, and requires no clarification is competing against a smaller field and a larger remaining balance than one that arrives at 11:59 p.m. on September 9.

The practical instruction is straightforward: if your package is complete, submit before the deadline rather than at it. If it is not complete, the pieces most likely to be missing are the ones that take the longest to obtain — signed cost-share commitment letters from partners, a demonstration site agreement, and the SAM.gov registration that has to be active and unexpired at submission. SAM registration renewal is the most common self-inflicted disqualification in federal grantmaking and it cannot be fixed on deadline day.

What to do with the next three weeks

If you are assembling a submission, four things separate a competitive package from a compliant one.

Lead with the demonstration, not the architecture. The NOFO asks for real-world data confirming a business case. Name the site, the operator or partner willing to host it, the traffic conditions, and the specific metrics you will publish. Reviewers evaluating "commercial viability" are looking for evidence you have already secured the environment where viability gets proven.

Quantify the opex claim. "AI-driven optimization reduces operational cost" is a sentence every applicant will write. A baseline, a target percentage, the measurement methodology, and the capital cost of getting there is a sentence very few will write.

Document the cost share as bindingly as you can. Letters that say a partner "intends to support" the project are worth less than letters specifying dollar amounts, in-kind valuation methodology, and signature authority. In a program where the match is half the project, the credibility of the match is half the proposal.

Address the exportability angle explicitly. NTIA has said out loud that it wants a trusted U.S. stack that allied nations will buy. If your solution has a plausible path into an allied carrier's network, say so, and say who you have talked to.

For organizations outside this program's narrow capital profile, the more useful takeaway is the pattern. The Innovation Fund's arc — authorized at $1.5 billion, rescinded by $850 million, redirected mid-stream by an administrator's speech, one full round canceled without awards — is a live demonstration of how much of federal grantmaking is now contingent on rescission bills and agency reprioritization rather than the authorizing statute. Applicants who built Round 3 proposals against a published NOFO learned that the hard way. Building a funding plan that assumes any single announced round will survive to award is no longer a safe assumption in any portfolio.

Applications close September 9, 2026, through the process described in NTIA's Round 4 program page, and the rolling award structure means the useful deadline is earlier than the official one. When you are tracking a program whose scope and balance can move between rounds, tools like Granted help you keep the current terms, deadlines, and eligibility in front of you instead of working from a NOFO that has quietly been superseded.