This appears to be changing at the National Science Foundation. Earlier this month, we detected the first outright termination of a NSF grant since last year. In addition, NSF is starting to use new mechanisms to disrupt research programs. As reported by Berkeleyside and Nature, 18 NSF grants to researchers at UC Berkeley were “suspended” in April, based on claims of insufficient disclosure of foreign funding. Researchers claim to be baffled as they have received no foreign funding. You can view these 18 grants now in our database. They span a wide range of topics, grant types, and directorates.

The May termination appears to target the kind of diversity topics the administration has been trying to suppress. The April Berkeley suspensions do not. Rather, they appear to use a new strategy to target UC Berkeley or the University of California, whose faculty has successfully fought back terminations in court, broadly and opportunistically. Two of the 18 suspended grants, in fact, were terminated in 2025, then restored by court order in Thakur v. Trump.

We need your help to continue to track these grant disruptions. “Suspensions” of the type experienced by UC Berkeley researchers do not appear in our data sources. Even if the suspensions are indefinite, they are not “final actions” and thus are not reported in public grant records. Other such disruptions may be occurring quietly in institutions across the country as the government tries new strategies to evade legal restrictions and oversight. NSF has not updated its public list of terminated grants since June 2025. If you are a researcher that has been affected by a new grant disruption, or know of one, please report it via our form, email, or Signal. Your reports help us track the administration’s actions and advocate for affected researchers and the public being denied the benefits of their science.

With a series of new obligations to continuing awards in April, we had to adjust the way we handle the integration of the more up-to-date research.gov data with more comprehensive USAspending data. Research.gov does not break up within-fiscal year obligations more granularly than the fiscal year, so we made an assumption that those obligations occurred at the most recent update of the grants. While this held true for the first half of the fiscal year and most grants that receive one obligation per year, some continuing awards, especially large infrastructure projects, received additional new obligations in April after previous ones in late 2025. Due to our assumptions the combined obligations showed up as a single recent spike in April, while in reality they occurred in multiple installments. Now the month-to-month fluctuations have been re-arranged and now accurately show the timing of obligations early in the fiscal year, and avoid any double-counting with USAspending data.

An example is can be seen in before-and-after plots of new obligations in the Geosciences Directorate. The first plot shows the old version of the curve, where the spike in April is visible. The second plot shows the updated curve. The difference is almost entirely due to the fact that the National Center for Atmospheric Research didn’t receive all its FY 2026 obligations in so far April, but split between December and April.

Thanks to Dan Garisto at Nature for flagging this issue.

As always, if you have any questions about our methodology or notice any issues in our data, please reach out at info@grant-witness.us.

As a result of this change, three reinstated supplements were added to our NIH data.

There are many reasons why a grant may recieve $0 in outlays in a particular period including that it is simply fully paid out. Because of that, one of our criteria for a frozen grant is that its total outlays didn’t match it’s total obligations. At the time of the freezes, this worked. However, as grants became unfrozen and received additional obligations and outlays, this logic stopped making sense as it was using the current total outlays and obligations. The fix was to check that a grant was not fully paid out at the time of institutional targeting which we can get by subtracting outlays/obligations since the start of the freeze from the current total outlays/obligations.

We also made some changes to better account for the fact that grants may move away from or to institutions while they are experiencing a freeze.

The results are that, compared to last week, 134 grants no longer meet the criteria for frozen/possibly unfrozen/unfrozen because they were paid out or nearly paid out at the time the institutional freeze began and therefore are no longer in our data. On the other hand, 63 grants were added to our data with the status of frozen, possibly unfrozen, or unfrozen because they were not actually fully paid out when a freeze was enacted.

There is still an outstanding issue with frozen grants at Duke University. Many grants at Duke were frozen in late July of 2025 and while most of these frozen Duke grants still have not received outlays, a good number of them appear to have received outlays in December 2025 (our most recent outlay data as of the time of writing). Currently our criteria require that a grant must have no outlays the entire time an intitutional freeze is in place to be frozen, and as far as we know it is still in place. Because these grants don’t meet the criteria, they have been removed from our data rather than being updated to “unfrozen” status (116 of the 134 grants removed are currently at Duke). We are currently working on investigating what is going on at Duke with frozen grants and coming up with a definition that would better suit the situation.

The GRFP has supported high-potential STEM graduate students since 1952. Over the past year the administration has waged a multi-front assault on the program: delaying the solicitation leaving only weeks to apply before deadline, narrowing eligibility, slashing the number of fellowships from 2,300 to 1,500, skewing the awards toward AI and quantum science, and rejecting applications without review.

The forward-funding shift is a tactic that the administration has used across multiple agencies to cut the number of awards while maintaining the appearance of spending the same amount of money. At the National Institutes of Health, a requirement that 50% of new awards be funded was imposed in the latter half of 2025, resulting in a 15% drop in the number of new awards according to NIH’s own analysis. The administration has also used forward funding to cut the number of awards at the Department of Energy’s Office of Science, and the Department of Defense’s research agencies.

This also shows how forward funding is being used beyond the National Institutes of Health. While the FY2026 appropriations bill put some guardrails on this strategy for NIH, it did not do so for NSF, perhaps because NSF has long used forward funding for the majority of its grants.

NIH itself describes the drop in new awards in 2025 as being to due to the increase in forward funding.

The 2022 CHIPS and Science Act reauthorized the GRFP with the target of supporting over 3,000 new fellows annually by 2026. Congress appropriated $325 million for the program in both FY2024 and FY2025. The FY2026 appropriation dropped to $285 million. With continued forward funding, we would expect only about 1,100 awards this year. The administration’s budget request was just $128 million - 60% below this the 2025 funding level and 76% below the CHIPS Act target. Combined with forward funding, this would likely have resulted in fewer than 500 new awards in 2026, fewer than any year since the program began in 1952. <–! Note, this is a bit of a stretch as the ongoing costs would be lowered. I could run some calculations to estimate this. –>

The table below shows how this works. In 2022–2024, NSF obligated roughly $110,000–$112,000 per new fellowship, consistent with funding one year of support at a time. In 2025, that figure jumped to $258,000 — more than double — even as the number of new awards dropped from over 2,000 to 1,500. Total new obligations actually rose to $387 million, the highest in the table, despite the program awarding the fewest fellows. Meanwhile, total outlays — the funds actually paid to universities — remained roughly flat around $215–$229 million across all years, reflecting that students draw down stipends at a steady rate regardless of how the obligations are structured.

This is not a matter of a change to who is recieving the awards or which institutions’ tuition is paid. The pattern holds at the institution level. For schools with enough GRFP history to compare over time, the amount obligated per fellowship in 2025 was 2.6 times greater than in 2022–2024. Obligations are for both new and continuing awards, but as large institutions typically have similar numbers of new and continuing fellows each year. Wwe would expect the obligation per new fellow to remain steady year-to-year. Instead, we see that the obligation per new fellow jumped dramatically in 2025 for each institution. This is consistent with a shift to forward funding for new awards in 2025.

NSF uses forward funding for many of its grants, and multi-year obligation can be a fiscally prudent strategy. However, the rapid shift to forward funding, as the administration has enacted here and elsewhere has the effect - and likely the intent - of cutting the number of awards while maintaining the appearance of spending the same amount of money. Even if, after a period of transition, forward-funding would cover the same number of awards, the interim period creates a permanent impact on scientific capacity, with fewer students starting careers in science.

For this reason, the administration reportedly fought hard in the FY2026 spending bill to retain the ability to use multi-year funding.

number of fellows:

• April 2025: NSF announced it would award only 1,000 fellowships, less than half the ~2,300 in prior years. After public outcry, 500 more were added — but tilted heavily toward AI and quantum science, shutting out the life sciences entirely.

• 2025: Former program officers Susan Brennan and Gisèle Muller-Parker document how the 2025 cycle shifted from awarding fellowships based on student potential to favoring specific fields aligned with the administration’s priorities.

• September 2025: NSF released the 2026 solicitation nearly two months late with dramatically narrowed eligibility. Second-year graduate students — previously told they could apply — were suddenly excluded. Students who had withdrawn 2024 applications expecting to reapply were locked out. When eligibility was last changed in 2016, NSF gave seven months’ notice and grandfathered affected students.

• February 2026: NSF began rejecting applications without peer review. Grant Witness has analyzed at least 50 such rejections, most involving proposed research in the life sciences — including proposals nearly identical to ones that earned honorable mentions the year before.

• FY2026 budget: The administration requested just $128 million for the GRFP — a 60% cut from FY2025 and 76% below the CHIPS Act authorization.

The GRFP is funded out of NSF’s STEM Education accounts, which also funds education research in the STEM Education Directorate, which was targeted for massive grant terminations in 2025.

The Trump administration plans to terminate $600 million in CDC grants to California, Minnesota, Colorado, and Illinois, as reported by the New York Times, the New York Post, and The Hill. A list of these grants has been circulated to Congress. We provide the full list below.

The vast majority of grants cut are operational funds to state and local health departments to support projects such as public health workforce recruiting, retention, and training, and to update and improve IT and data infrastructure.

The administration also circulated a set of talking point examples of grants to be terminated, cherry-picking the most luridly “DEI” language from grant descriptions, titles, and funding programs. You can read the full examples here. The language from these talking points has been reproduced in much of the press coverage. These examples total $33 million in funds to be cut, or approximately 5.5% of the $600 million total.

It is unclear whether all of the talking points refer to grants on the congressional list. For example, a grant to Colorado to “Address COVID-19 Health Disparities Among Populations at High-Risk and Underserved, Including Racial and Ethnic Minority Populations and Rural Communities” was not on the list of grants to be terminated shared with Congress.

Many of these grants are Public Health Infrastructure Grants (PHIGs). The administration “paused” then restored this $5B national program over 24 hours on January 23-24. All but 5 PHIGs in these four states are on the list of grants to be terminated.

While we don’t yet know why some PHIGs may survive while most others are terminated, it’s notable that 4 of the surviving 5 PHIGs are in “red” counties. In Colorado, the only surviving PHIG is to El Paso County, where 54% of voters cast their ballot for Trump in 2024. El Paso County is also represented by a Republican member of Congress (Jeff Crank, CO-5) at the moment. In California, 2 surviving PHIGs are in Riverside and San Bernardino Counties, which also voted for Trump over Harris in 2024. A 3rd surviving in PHIG in California is in Orange County, a traditional Republican stronghold and home to an outsized proportion of wealthy Trump donors. The final surviving PHIG in California is in Sacramento County, and it is the only surviving PHIG outside a Republican stronghold.

Grant Witness will be launching a CDC grant termination tracker shortly. If you have information about CDC grants being terminated, please fill out our CDC reporting form.

In this post, we dive deeper into the types of grants that were targeted on January 13th, investigating trends of disrupted program assistance listings, program offices, and notices of funding opportunities for these grants.

To begin, we broke down our dataset on SAMHSA grant disruptions based on whether the grant had been specifically targeted on January 13th, since the dataset includes information on disrupted grants as early as March of 2025.

Note: the majority of grants targeted on January 13th have been reinstated, so the listed “number of grants” and “disrupted funding” in the following tables and visualizations do not represent the current loss from SAMHSA grants (i.e. the loss from SAMHSA grants that are still terminated). To see the most up to date information on the amount lost in SAMHSA grants and funding, view our dataset at SAMHSA grant disruptions.

1. They were non-renewed grants—the project’s budget year ended and the PI did not receive an expected notice of award for the subsequent budget year. Some of these grants may have never received a renewal, effectively ending the project, but this is a type of grant termination we don’t currently track (we are working on it though!).

2. PIs may have felt their grants were very likely to be terminated and submitted them as part of a lawsuit despite not having formally received a termination letter.

We now feel it is best that we remove any grants from the public data that were in these court documents but don’t have other signals of termination unless they were at institutions known to have been targeted by the Trump administration.

As always, if you have questions or notice any issues with our data, please let us know at info@grant-witness.us.

As noted previously, correctly determining the current status of grants has been tricky since grants can go through multiple rounds of terminations and reinstatements To that end, we have relied on combinations of event histories to determine the status of a grant.

In both the data table and the CSV file, you’ll find an event_history column with a bulleted list of all the relevant “events” for a grant that have led to its current status. Similarly, there is now a reinstated_indicator alongside the termination_indicator column which specifies how a grant is identified as terminated or reinstated.

For a grant to be listed as “🔄 Possibly Reinstated”, it must either:

1. Have a project date re-extension (as identified through USASpending, NGGS or File C data), or
2. Was manually reported as reinstated (i.e. via court order as in the cases Sustainability Institute v Trump and Green & Healthy Homes Initiative, Inc. v. Environmental Protection Agency)

The exception would be if the grant has been terminated again, in which case the status would be updated and the event_history column would note this history.

Any grant that meets any of the conditions for “🔄 Possibly Reinstated” AND has had an outlay since its reinstatement is marked as “✅ Reinstated”.

While date re-extensions, obligations and outlays for grants are continuously monitored and updated through our data pipeline, court reinstatements are identified manually.

As always, if you have any questions about our methodology or notice any issues/mistakes in our data, please feel free to reach out at info@grant-witness.us.

Introduction

Late in the evening on January 13th, 2026, with no warning or prior notice, the Trump Administration had cancelled $2 billion in funds to 2,700 separate mental health programs in all 50 states.

These programs provided crucial support to a wide range of Americans, from veterans and healthcare workers to grade school students and new mothers. These services included mental health treatment, prevention, education, and public awareness campaigns. Some grants also sustained training programs for first responders. For example:

• 417 programs aimed to prevent suicide, 465 mentioned alcohol use, and 351 helped Americans living with HIV.
• Another 218 grants helped Americans with depression, while 9 supported adults living with schizophrenia and their caregivers.
• 573 grants focused on services in rural areas.
• Over 300 grants supported law enforcement, who are often called on to help Americans in mental health crises.

According to standardized form letters that Trump officials sent to each mental health organization, these programs were cancelled because the Trump Administration no longer wanted to support them. Trump officials wrote that the programs “no longer effectuate” their “goals or … priorities.”

This boilerplate wording is identical to language that Trump officials used in prior mass grant cancellations at NIH, NSF, and EPA. In many cases, Federal courts have found those cancellations to be unlawful.

The blow back from these cancellations was immediate and overwhelming. Advocates, service providers, state officials, and ultimately members of Congress fought back–and ultimately won, at least for now. Less than 24 hours later, Trump officials backtracked and said they would rescind the terminations and restore these programs.

Starting today, we added University of Pennyslvania to our list of institutions that had been targeted. Funding freezes at Penn began in March and ended in July.

We now distinguish between “full freezes” where all grants at an institution are known to be frozen (i.e. Columbia and Brown), and “partial freezes”. When there is a “full freeze” our rules for detecting a frozen grant are a bit relaxed—we don’t require $0 outlays the entire targeting period, rather all grants at that institution are considered frozen after targeting begins and are considered “unfrozen” if they receive outlays, even if this is before targeting of the institution ends. This change is important, since it appears that some grants at Columbia recieved outlays before the university officially capitulated.

We also made some changes to better handle grants that move from targeted institutions during freezes and to handle periods with “expected” $0 outlays due to a grant being nearly or fully spent and near its project end date.

The result is the addition of 64 “possibly unfrozen”, 160 “unfrozen” grants, and 153 “frozen” grants and the removal of 44 “possibly reinstated” and 10 “frozen” grants. The 44 “possibly reinstated” are all currently at Brown and Columbia, and were incorrectly marked “possibly reinstated” for a while now due to a mistake on our part as these grants were never terminated, only marked as frozen under the earliest definition we used and are no longer detected as frozen. Six of the eight “frozen” grants removed are at Duke and are removed due to an incorrect freeze start date of April being used prior to this update. The 153 “frozen” grants added are also all at Duke and weren’t showing up previously due to a bug.

We’ve set aside the full list grants that were removed by these changes to investigate further and confirm what really happened to them. If you have a grant that had frozen funds, and it’s not in our data, please contact us at info@grant-witness.us!

There may still be issues with the status of NIH grants that have complicated histories (e.g. frozen, terminated, reinstated, then unfrozen!) that we will continue to investigate. If you spot any mistakes or inconsistencies in our data, please let us know by email.

A grant that meets any of these four conditions is marked as “❌ Terminated”:

1. It was self-reported by the PI with a termination letter.
2. It was reported terminated by DOGE and has had it’s project end date moved earlier in either NGGS or USAspending (a date “cutoff”).
3. It was part of a program that has been terminated (as reported by a verified source) and has had either a date cutoff or a deobligation of funds with no further obligations.
4. It is on the NCIF/CCIA list of terminated grants or on the Our Environment and Public Works Senate Commimttee list of terminated grants.

In the data table and the CSV file, you’ll find an event_history_public column with a bulleted list of all the relevant “events” as well as a termination_indicator column indicating how we come to the decision that a grant has been terminated.

For the website’s data table, there are tabs that filter grants by the impacted EPA programs, including:

• Environmental Justice Programs (66.306, 66.309, 6.312, 66.604, 66.614, 66.615, 66.616) - These grants focused on supporting and empowering communities that work on solutions to local environmental and public health issues.
• STAR: Science to Achieve Results (66.509) - Grants under this program fund competitive, peer-reviewed research on environmental and public health issues.
• Solar for All (66.959): The goal of this program was to make solar energy accessible and affordable for low-income and disadvantaged communities.
• CCIA: Clean Communities Investment Accelerator (66.960) - Part of the Greenhouse Gas Reduction fund, this program provided clean energy funds and technical assistance to underserved and disdavantaged neighborhoods
• NCIF: National Clean Investment Fund (66.957) - Part of the Greenhouse Gas Reduction fund, this program created national clean financing institutions to provide loans, loan guarantees, and other financial products to help accelerate the transition to clean energy.

Importantly, we have not yet included reinstated grants in our data. We know there have been reinstated EPA grants, but correctly determining their current status is tricky since grants can go through multiple rounds of termination and reinstatement. We are working on implementing this and will post an update once reinstatements are accounted for.

As always, if you notice any issues or mistakes in our data, please let us know at info@grant-witness.us.

Candidates for frozen grants must meet the following criteria:

1. At an institution that has been targeted for freezes
2. Has a project end date after freezes began
3. Between the month after the month in which the targeting began and the month before the month in which the targeting ended, received only outlays < $100.
4. Has had total outlays of at least $100 in the 6 months prior to the start of targeting
5. At least 20% of the periods prior to Jan 2025 had positive outlays (this excludes grants that have no info in File C before Jan 2025 as well)
6. Aren’t 100% outlaid (i.e. the total outlaid is not equal to the total award amount)
7. If they are within 3 periods of the project end, aren’t more than 95% outlaid

Finally, if a frozen grant was since terminated, it is listed as terminated.

This definition is slightly different from the original definition we used and while we believe it is still quite conservative, it captured an additional ~400 frozen grants that were not previously in our initial attempt in June (we also added Duke University to the list of targeted institutions since then).

“Possibly Unfrozen” grants are grants that met the above definition at some point, but are no longer known to be targeted (e.g. at Columbia and Brown, due to capitulation to White House demands). “Unfrozen” grants are “possibly unfrozen” grants that have received outlays after targeting ended. Frozen grants that reached their project end date before receiving outlays remain listed as “frozen” or “possibly unfrozen”.

In the spirit of transparency, in the downloadable CSV, frozen grants have a file_c_outlays column listing the history of outlays from usaspending File C and targeted_start_date and targeted_end_date columns indicating the time the institution was targeted for freezing. You can also now see the outlay history for frozen grants in the individual record view by clicking on a row in the NIH table.

If you spot any frozen grants that you don’t think should be marked as such, please don’t hesitate to contact us.

There were 211 grants listed in these cases that were not previously reported to have been terminated – we’ve now added these to our database. We have not included subawards that were part of Massachusetts v. RFK Jr. nor grant renewals that have been delayed and were part of the case.

Grants restored via these cases have “🔄 Possibly Reinstated” tags, with references and links to the relevant court documents in the new court_restoration_url column. We list these as possibly reinstated because we do not yet know what the process of reinstatement looks like or how long it will take. If your grant has been reinstated, please help us understand what’s happening by filling out our submission form!

We’ve added other visual tags – “❌ Terminated” and “🧊 Frozen Funding” – to help distinguish between the various types of suspended and cancelled research. There are also new tabs that allow you to separately browse ever terminated, possibly reinstated, and frozen grants.