A New Database Says $177 Billion in Grants Was Cut or Frozen. The Third Category It Tracks Is the One That Should Change How You Pick Opportunities.
September 16, 2026 · 7 min read
Granted Research Team · Editorial policy
On Wednesday, September 16, 2026, the States United Democracy Center and a group called Grant Witness published Lost Funds, a public database at lostfunds.org that attempts to count every federal grant disruption since January 2025 in one place. The headline number is $177 billion — described by both organizations and by the Associated Press as an up to figure, which matters — and it works out to roughly 10 percent of federal discretionary spending.
That number will be quoted in a great many board decks over the next month. Most of the people quoting it will not have opened the database, and will therefore miss the thing inside it that is actually operationally useful.
Lost Funds does not track one phenomenon. It tracks three, and it says so explicitly in its methodology:
- Formal cancellations of committed grants and contracts
- Freezes that prevent a recipient from drawing down money it has already been awarded
- Slowdowns in the rate at which grants are issued
Category one is what makes the news. Category two is what fills courtrooms. Category three is nearly invisible, cannot really be litigated, and is the category that determines whether the notice of funding opportunity sitting in your tab is worth the eighty hours it will cost you.
What is actually in the database
The structural detail is more interesting than the topline. According to the project's own summary, the disruptions touch:
| Recipient type | Count |
|---|---|
| Local governments | 3,494 |
| School districts | 753 |
| Businesses | 437 |
| Other recipients (nonprofits, universities, tribes, health systems) | 15,424 |
| Total | ~20,108 |
Geographic coverage is all 50 states plus the District of Columbia. California, Texas, New York, Illinois, and North Carolina carry the largest disrupted dollar totals — which is mostly a population-and-institutional-density artifact and should not be read as targeting. The largest sector shares are health, nutrition, environment, and disaster relief. The AP's illustrative examples span the political geography deliberately: maternal health funding in Michigan, education research in Mississippi, assistance to minority farmers in Iowa.
The litigation column is the genuinely surprising one. States have joined 23 lawsuits, and the project credits those suits with restoring $35 billion in disrupted funding. That is not a rounding error. It is roughly a fifth of the gross disruption figure, and it means the honest framing of the headline is closer to $177 billion disrupted, $35 billion of it clawed back so far, the remainder in varying states of contested.
On methodology: the primary source for award amounts is USAspending.gov, refreshed monthly. The identification of which awards were disrupted comes from a messier blend — agency communications, administration directives, media coverage, litigation filings, and direct submissions from affected recipients. Enhancements land biweekly. The whole dataset downloads as CSV, per state, free, with a suggested citation format.
That last detail is the one that turns a news story into a tool.
Why "up to" is doing real work in that sentence
If you are going to cite this number in a grant application, a board memo, or a legislative testimony, cite it accurately, because the construction is load-bearing.
A canceled award and a slowed award are not the same liability. An award that was frozen in March and unfrozen by court order in July appears in the disruption count and also in the $35 billion restored — the project is transparent about tracking both, but a careless reader will double-count. And "slowdowns in the rate at which grants are issued" is an estimate of money that would have flowed under a prior issuance pace, not money that was pulled out of a specific recipient's account.
None of that makes the figure soft. It makes it a range whose components behave differently, and the components are separable in the CSV. Anyone building a case on this data should separate them before publishing. If you assert "$177 billion was taken from communities" in a proposal narrative and a program officer who has read the methodology is reviewing it, you have spent credibility you did not need to spend. "Lost Funds documents roughly 20,000 affected recipients and up to $177 billion in disrupted federal funding, of which about $35 billion has been restored through litigation" is both more accurate and, frankly, more alarming.
The slowdown category is the one that changes your pipeline
Here is the analytical core.
If a program cancels your award, you know immediately. You get a letter, you get a termination clause citation, and — post-Talwani and the related rulings on the agency-priorities termination clause at 2 CFR 200.340 — you may well get a lawyer. The harm is legible, dated, and appealable.
If a program simply issues fewer awards, nothing happens to you at all. There is no letter. There is no defendant. The NOFO still posts. The eligibility language is unchanged. The anticipated-awards line in the funding opportunity may even read the same as last year, because that line is an estimate and agencies are under no obligation to hit it. You write the application, you wait five months, and you get a decline that looks exactly like an ordinary competitive decline.
Multiply that across a sector and you get what practitioners have been describing anecdotally for eighteen months — the sense that everything is harder without any single identifiable thing having gone wrong. Lost Funds is the first broadly accessible dataset that tries to quantify that specific feeling.
For an applicant, the practical consequence is that program-level issuance velocity is now a first-class selection criterion, ranking alongside eligibility fit and competitive positioning. It was always a factor. It used to be a factor you could safely ignore, because federal issuance rates were boringly stable. They are not boringly stable now.
A concrete diligence routine
The CSV download makes this cheap. Before committing serious effort to a federal opportunity, run four checks:
1. Has this specific program appeared in the disruption data? Search the state CSVs for the program name and the assistance listing number. A program that shows cancellations is a program whose awards carry termination risk even if you win. A program that shows nothing is not necessarily safe, but it is at least not on the board.
2. Is the disruption a freeze, a cancellation, or a slowdown? Freezes are the most recoverable — courts have repeatedly ordered funds released, and the $35 billion restored figure is largely a freeze-and-condition story. Cancellations are worse but at least discrete. Slowdowns mean the odds are quietly worse than the NOFO's anticipated-awards line implies, and no amount of proposal quality corrects for a denominator that shrank.
3. Is there active litigation, and is your program in it? The database links case documents through CourtListener. If your program is inside a suit that has already produced an injunction, the practical funding environment may be better than the news coverage suggests. The HUD Continuum of Care competition is the cleanest example of a court materially changing the state of a live competition — twice, in consecutive years.
4. What does the recipient-type mix tell you about your own class? If 753 school districts appear and you are a school district, that is a base rate. If 437 businesses appear and you are a small business pursuing SBIR, that base rate is much lower and your exposure profile is different. Do not apply sector-wide anxiety to a lane that the data says has been comparatively stable.
That routine costs perhaps ninety minutes per opportunity and it will, over a year, redirect real effort away from applications that were never going to be funded at the historical rate.
What the data does not tell you, and what to do about it
Three real limits.
USAspending lags. The transaction data refreshes monthly and reflects obligations, not intent. A program that quietly stopped issuing awards in August will not be legible in the numbers until well into the fall. The absence of a signal is weak evidence.
Slowdown attribution is inferential. Distinguishing a policy-driven issuance slowdown from an ordinary appropriations-timing artifact is genuinely hard, especially in a year running on a continuing resolution that funds agencies only through December 11, 2026. Some of what reads as suppression is a CR doing what CRs do.
The regulatory baseline is itself unsettled. The OMB rewrite of 2 CFR Part 200 that was scheduled to take effect October 1 is statutorily blocked through December 11. If it lands in the new year, the termination and pre-award review provisions change the shape of future disruptions, and a database calibrated to 2025–2026 conditions will need recalibrating.
The organizations behind the project are open about their posture. States United describes itself as nonpartisan and focused on rule of law and elections; it was cofounded by Norm Eisen, who left in 2024. Grant Witness is a group of scientists, researchers, and attorneys formed to document funding changes. Kelly Rader, States United's research director, framed the launch around scale and state response. Scott Delaney, Grant Witness cofounder, put it this way: "By bringing thousands of funding disruptions together in a publicly accessible, verified database, Lost Funds displays their impact." The White House did not immediately respond to the AP's request for comment.
You do not have to share the project's framing to use its CSVs. The underlying award records come from the government's own spending database, and they are checkable line by line.
The strategic read
The instinct after a report like this is diversification, and diversification is correct — but it is generic advice that most organizations already know and few execute, because diversifying a funding base takes two years and the crisis is now.
The narrower, more actionable takeaway is this: the disruption is unevenly distributed across programs in ways that were previously unknowable and are now, for the first time, downloadable. Twenty thousand recipients is a large enough sample to see structure. Some programs in your portfolio are carrying far more risk than their sister programs at the same agency, and you have been treating them as interchangeable because you had no basis not to.
You have a basis now. Use it before you pick the next thing to write.
Sources: Lost Funds project · Associated Press via ABC News · Anchorage Daily News · The Boston Globe