Eight Days Left to Obligate: NSF Is $1 Billion Behind, AHRQ Will Strand $160M, and a Judge Wants a Status Report by September 28

The federal fiscal year ends Tuesday, September 30, 2026. Eight days from today. For most grantseekers that date is background noise. This year it is the story, because several federal science agencies are going to hit it holding money they were appropriated and did not award.

Inside Higher Ed reported the numbers on September 21. They are worse than the general sense of "things are slow" suggests.

The scoreboard, eight days out

National Science Foundation. As of September 12, NSF had obligated $6.3 billion across roughly 6,200 grants. At the same point last fiscal year: $8.1 billion across 8,600 grants. That is a gap of more than $1 billion and roughly 2,400 awards. Worse, the pace has not merely slowed — grant issuance has been essentially flat since around August 20.

Scott Delaney, co-founder of Grant Witness, which has been tracking agency obligation rates all year, put it plainly: "There are probably more science agencies that are struggling than are succeeding in hitting their targets." His sources inside NSF say the agency is effectively finished making grants for the year and will not come close to prior-year output.

Agency for Healthcare Research and Quality. As of September 13, AHRQ had made 61 awards totaling $27 million. Last year at the same point: more than 500 projects and $181 million. AHRQ is on track to leave roughly $160 million unspent.

That is not a slowdown. On dollars, that is an 85% collapse. AHRQ's entire extramural research enterprise — the patient-safety work, the health-systems research, the primary-care practice-based networks — ran at one-seventh of its normal volume this year.

Institute of Education Sciences. Of the $768 million Congress appropriated for fiscal 2025 education research — money that expires September 30, 2026 — IES had spent only $587 million as of August 26. Roughly $180 million remains unobligated by the most recent plaintiff accounting, and earlier estimates ran as high as $289 million. The breakdown is lopsided: special education research was about 85% unspent ($65.5 million of $77 million), other education research more than 50% unspent, statistics funding 40% unspent. NAEP, by contrast, was fully funded.

For historical calibration: under prior administrations IES typically obligated at least 90% of a two-year appropriation by August 1 of the second year. During the first Trump administration, the agency never had more than 3% left at this point in the cycle.

The Burroughs order

On Thursday, September 19, U.S. District Judge Allison D. Burroughs of the District of Massachusetts ordered the government to file a status update by September 28 — two days before expiration — on whether IES will in fact obligate the remaining funds.

Her language is worth reading as written:

If it appears, as Sept. 30 approaches, that Defendants have misled the Court ... the Court may reconsider whether there has been undue delay warranting injunctive relief.

That is a judge telling an agency she is aware the calendar itself can function as a merits ruling. If an agency represents in court that it intends to obligate funds, and the clock runs out, the money is gone regardless of how the litigation is eventually decided. Burroughs is establishing a record on that point before the deadline rather than after it — and reserving the option to act.

The Education Department's position is that IES intends to use the remaining funds "to fully fund existing research grants, as well as to support other high-quality statistical and evaluation activities." Note what that formulation does: fully funding existing grants obligates money without running any new competition. It is the fastest legal way to spend down a balance in eight days. It is also, from a field-building perspective, the least generative — no new investigators, no new questions, no new awards.

Beyond the fiscal 2025 money, more than $700 million in IES fiscal 2026 funds remains largely unobligated, with roughly $465 million of it withheld at OMB. We covered the restart attempt in IES Restarts FY27 Grantmaking.

What actually happens on October 1

There is widespread confusion about this, and the mechanics matter for how you should read the next three months.

For most agency program accounts, unobligated balances of a one-year appropriation expire at the end of the fiscal year. The budget authority lapses. The agency can no longer make new awards against it. This is the "use it or lose it" rule people invoke loosely, and for the accounts in question here it is literally correct — IES funds that are not obligated by September 30 return to the Treasury.

A few clarifications that are routinely mangled:

• Expiration is not cancellation. An expired account stays open for five more fiscal years to handle adjustments to obligations already made. It is only at the end of that fifth year that the account is closed and all remaining balances — obligated and unobligated alike — are cancelled. But you cannot make new awards from an expired account.
• This is not the same thing as grantee carryover. If you are a current recipient with unspent funds in your budget period, carryover is a separate mechanism requiring either automatic authority or Grants Management Officer approval. Agency-side expiration does not touch it.
• Multi-year and no-year accounts behave differently. Not every dollar in these agencies is one-year money, which is why headline "unspent" figures and "will expire" figures are not always the same number.

The practical consequence: September 30 is a real, irreversible boundary for a specific slice of this money, and a soft one for the rest. Reporting that treats all unspent federal research dollars as vanishing on Tuesday is overstating it. Reporting that treats none of it as vanishing is understating it.

Why NSF is behind, and why that is the most important number here

NSF's shortfall has a different cause than IES's, and the difference should shape how you plan FY2027.

IES is a capacity-and-approval story: the agency lost more than 160 of its 190 staff, and OMB has been slow to release apportioned funds. Those are correctable conditions, at least in principle.

NSF's is a structural reallocation. More than $1 billion of the agency's roughly $8.8 billion budget is held in a central account that core grant-issuing programs cannot reach, much of it earmarked for the White House's Grand Research Challenges initiative targeting artificial intelligence and advanced materials. Add roughly 400 peer-reviewed proposals stuck in processing at the Office of Award Management, plus workforce reductions, and you arrive at a projected ~6,100 new grants for FY2026 — about 46% below the 2021–2024 average and the lowest count in more than four decades.

That distinction is the whole planning insight. NSF did not fail to spend $1 billion. NSF spent it somewhere else — or is about to. The money is being routed from thousands of investigator-initiated awards into a small number of large, directed, translation-oriented instruments: X-Labs, the Tech Accelerators, ELEGANT, Grand Research Challenges. We laid out the mechanics in NSF's $1 Billion Central Account.

An IES-style shortfall might reverse next year. An NSF-style reallocation is a new baseline until an appropriations cycle reverses it, and nothing in the current one does.

The cross-agency pattern

Put the three together with NIH, whose own year-end compression we covered in NIH Is Sitting on More Than $1 Billion It Must Obligate by September 30, and the shape is consistent:

1. Competitions ran late or not at all. Fewer NOFOs, later postings, shorter windows.
2. Review-to-award timelines stretched. The standard federal research grant cycle is six to ten months from solicitation to award. Compressing that into a fiscal year's final quarter is not possible, which is why an agency that is behind in June is mathematically behind in September.
3. Spend-down favors continuations over new competitions. When an agency must obligate fast, funding existing awards is the only reliable lever. New investigators lose disproportionately.
4. Approval bottlenecks sit above the agency. OMB apportionment, not program office capacity, is the binding constraint in several of these cases.

Point 3 is the one to internalize. In a compressed year, the distribution of pain is not uniform. Established labs with continuing awards get made whole. First-time applicants and early-career investigators — who by definition have no continuation to be folded into — absorb almost all of the shortfall.

What to do with this

If you have a proposal pending at NSF right now. It is very likely not getting a FY2026 decision. Roughly 400 proposals are in processing limbo, and issuance has been flat for a month. Ask your program officer directly whether your proposal is in the FY2027 queue. That is a fair, answerable question and the answer changes your fall.

If you were planning to submit to AHRQ. Plan around the agency, not through it. Sixty-one awards is not a funding program you can build a lab budget on. The nearest substitutes for health services and patient-safety work are NIH institutes with overlapping portfolios, PCORI, and the larger health conversion foundations. Start those conversations now rather than waiting to see whether AHRQ's FY2027 looks different.

If you are an education researcher. Watch the September 28 filing in the Burroughs case. It is the clearest available signal on whether IES's fiscal 2026 money — the $700 million-plus, expiring in September 2027 — moves on a normal schedule or repeats this year. That filing tells you whether to build a fiscal 2027 submission plan around IES at all.

If you are building an FY2027 strategy generally. Three things are simultaneously true and you should plan for all of them:

• The current continuing resolution funds the government only through December 11, 2026, which is also the date OMB's Uniform Grants Regulation rewrite becomes finalizable again. We covered that convergence in The Reprieve Has a Date on It. Q1 FY2027 is not a stable planning environment.
• Money at NSF is not shrinking so much as concentrating. Large directed instruments are growing while core programs contract. If your work can be framed against a translation endpoint with an industrial or mission partner, the large instruments are where the dollars went.
• Diversification is no longer defensive, it is baseline. A portfolio that is 80% single-federal-agency has a variance profile this year that it did not have three years ago. We walked through the rebalancing math in The 2026 Federal Research Funding Contraction.

The thing worth saying plainly

There is a version of this story that is purely about process — apportionment timing, staffing, court schedules. That version undersells it.

An appropriation is Congress deciding what research the country will buy. When $160 million of that decision expires at AHRQ and $180 million expires at IES, no one voted to stop buying patient-safety research or special education research. The purchase simply did not happen, and the deadline made it permanent. Judge Burroughs' order is notable precisely because she named that mechanism out loud: a deadline can decide a case that was never argued.

Eight days. Watch the September 28 filing.

Granted tracks federal obligation rates, agency award pace, and 400,000+ live funding opportunities. Search federal grants or start a free account to get matched to what is actually being awarded.