NIH Spent Every Dollar of Its $34 Billion and Still Funded 5,700 Fewer Projects Than Baseline. The Gap Is Not a Budget Story — It's an Obligation-Structure Story.
October 2, 2026 · 7 min read
Granted Research Team · Editorial policy
The headline from NIH's fiscal-year close is the kind of thing an agency press office likes: the full external grants budget — roughly $34 billion — went out the door. Nothing lapsed. Nothing was returned to Treasury.
The second number is the one that matters to anyone with a proposal in the queue. NIH funded approximately 58,000 projects in FY2026. That is down from about 61,000 in FY2025, and down from roughly 63,700 in the years preceding the current administration, according to the analysis published October 2 by STAT. Counts may tick up slightly as reporting databases catch up, but the direction is settled.
Put the two facts side by side and you get a conclusion that most lab-level planning has not yet absorbed: the same money bought roughly 5,700 fewer projects than the pre-2025 baseline. That is a 8.9% decline in funded projects against a flat-to-nominal budget. The constraint was never the top line. It was the shape of the obligation.
The arithmetic, stated plainly
Divide the external grants budget by funded projects and the mechanism becomes visible.
At $34 billion across 58,000 projects, the average obligation per project is roughly $586,000. Spread that same $34 billion across the pre-2025 baseline of 63,700 projects and the average falls to roughly $534,000 — about $52,000 less per project.
These are crude averages across wildly heterogeneous mechanisms, and they should not be read as a change in the size of a typical R01 budget. What they describe is something different and more consequential: NIH is obligating more dollars per award decision than it used to. The dollars are not smaller. The number of decisions is.
The reason is forward funding.
What forward funding actually does to a fiscal year
Under conventional annual funding, a five-year R01 draws roughly one-fifth of its total cost from each of five successive appropriations. Year one obligates year one. The out-years are commitments, honored from future appropriations, and they leave most of the current year's appropriation available for new competing awards.
Under multiyear lump-sum funding — also called forward funding — NIH obligates the entire project period at the time of the initial award, out of a single fiscal year's appropriation. That same five-year R01 now consumes five years of its cost from one year's money.
The accounting is clean. The consequence is not. Every dollar forward-funded into an out-year is a dollar unavailable to fund a new award this year. Scale that across thousands of competing research project grants and you get exactly what FY2025 and FY2026 produced: a fully obligated budget, a shrinking number of funded projects, and a cohort of applicants who scored well and were not funded.
STAT's reporting attributes the FY2026 decline directly to this: the smaller number of grants "potentially resulted from the White House's push to fully fund more multiyear grants in advance, which meant that there was less funding on hand to issue new awards."
Congress saw it coming and the count still fell
This is the part of the story that should change how you plan.
Congress did not miss the problem. The Consolidated Appropriations Act, 2026 (Public Law 119-75), enacted in February 2026, included a provision in Division B capping the amount NIH may obligate for multiyear awards at the level obligated for multiyear awards in FY2025. The accompanying explanatory statement recorded "strong concern about the impact of this policy on application success rates and the consequent reduction in the number of grants NIH can fund."
So the legislative branch identified the mechanism, wrote a cap, and attached report language naming the exact harm. And FY2026 still closed roughly 3,000 projects below FY2025.
Two readings are available, and they are not mutually exclusive. The first is timing: the cap arrived in February, nearly five months into the fiscal year, after a substantial share of FY2026 obligation decisions had already been shaped. The second is that a cap holding multiyear obligations at the FY2025 level locks in the FY2025 problem rather than unwinding it. FY2025 was itself a depressed year — 61,000 projects against a 63,700 baseline. Freezing the mechanism at that level preserves the squeeze; it does not release it.
Either way, the practical lesson is the same. A statutory cap on the mechanism did not restore the project count. If your FY2027 planning assumes congressional attention will translate into more funded awards, that assumption has now been tested once and failed.
Who absorbed the decline
A shrinking number of awards does not distribute its pain evenly. It concentrates on applicants without an existing portfolio to renew.
The early-stage investigator data is stark. Success rates for early-stage investigators seeking R01-equivalent grants fell from 29.8% in FY2023 to 18.5% in FY2025. The number of ESIs actually receiving awards dropped from more than 1,400 in 2024 to 1,144 in 2025. NIH's own extramural data commentary attributes part of the FY25 ESI decline — a funding rate of 18.9% — explicitly to NIH "forward funding more competing research project grants."
That is an agency naming its own mechanism as a cause.
FY2026 did not reverse it. STAT's analysis found that support for early-career scientists plateaued rather than expanding, despite expansion being a stated administration priority. The same analysis found little progress on another stated priority — distributing funds more evenly across the country — while research involving race, health disparities, and ethnicity absorbed further cuts, continuing the FY2025 trend.
The pattern across all three: the stated priorities that required more awards did not materialize, because the obligation structure was producing fewer awards. The priority that required fewer awards of a particular kind was achieved.
The delay problem is separate, and it compounds
Volume is only half of what researchers reported. The other half is timing.
In STAT's survey and interviews, researchers described delays in the grant review process slowing the pace of their work and making it harder to propose and conduct the ambitious, higher-risk science NIH exists to support. This is a distinct failure from the award-count decline. A late award and a missing award have different remedies, and conflating them produces bad strategy.
The self-censorship effect deserves particular attention. When review timing becomes unpredictable, rational applicants shift toward proposals that survive a long and opaque queue — incremental aims, established methods, preliminary data already in hand. The portfolio-level result is a quieter, less innovative applicant pool, which no funding-rate table will show.
What this means for an FY2027 submission
NIH enters FY2027 under a continuing resolution. Public Law 119-103, signed September 2, 2026, funds the government through December 11, 2026 at the FY2026 enacted level. And NIH has already told Institutes they may issue non-competing awards below the level printed on your most recent Notice of Award, with upward adjustments only to be "considered" after full-year appropriations arrive — the subject of NOT-OD-26-131, which we analyzed here.
Four adjustments follow from the FY2026 close.
Treat the award count, not the budget line, as your denominator. Advocacy and institutional planning that tracks the NIH top line will keep reporting good news while your odds get worse. The relevant metric is projects funded, and it has fallen in two consecutive years against a flat budget. Ask your research office for award counts by mechanism, not appropriations totals.
Assume out-year commitments are soft in both directions. Forward funding made some awards fully obligated and therefore unusually secure. The CR notice made non-competing continuations explicitly reducible. You may be holding either kind. Read the Notice of Award to determine which — a fully forward-funded award and an annually funded continuation now carry materially different risk, and the document is the only place that distinction appears.
If you are an ESI, the larger-award advice is real but conditional. NIH's September data commentary nudges early-stage investigators toward R01-equivalents rather than R21s, and the funding-rate tables support it. But that advice was written against a portfolio where fewer total awards are being made. A higher per-mechanism rate on a smaller base of decisions is still a smaller base of decisions. Pair the mechanism choice with a genuine assessment of review timing risk for your study section.
Build the delay into the budget, not just the calendar. The practical cost of a slow award is bridge funding, deferred hires, and lapsed effort commitments. Institutions that treated FY2025 delays as a one-time anomaly spent FY2026 learning otherwise. A third consecutive year is now the base case, not the pessimistic case, and the time to negotiate bridge terms is before the award is late.
The structural point
The FY2026 close is the cleanest natural experiment the extramural community has had in years. The budget was flat and fully spent. Congress identified the distorting mechanism and capped it. The project count fell anyway.
That tells you the decline is not primarily a function of how much money Congress appropriates. It is a function of how an agency chooses to obligate it — a decision made inside NIH and the White House, largely invisible in appropriations debate, and far less responsive to report language than anyone hoped.
For the FY2027 cycle, plan against the obligation structure rather than the top line. It is the variable that has actually been moving.
Sources: