NSF in June 2026: \$8.75 Billion Appropriated, 1,752 Grants Terminated by DOGE, Merit Review Cut From Three Reviewers to Two, and Why the Agency That Funds American Science Now Operates on a Fundamentally Different Risk Model.

The National Science Foundation that exists in June 2026 is not the agency that existed in January 2025. The budget survived Congress: $8.75 billion in FY2026 appropriations, almost exactly the FY2024 baseline, after the administration's proposed cut to $3.9 billion was rejected on the floor. But the agency that received those appropriations is structurally different from the one that built the modern American university research portfolio. Between April and May 2025, the Department of Government Efficiency terminated 1,752 active NSF grants worth $1.4 billion. The director resigned on April 24. The peer review process — the institutional core of NSF for sixty years — was rewritten effective December 15, 2025, dropping the minimum number of external reviewers from three to two, allowing one of those reviews to be conducted internally by NSF staff, making panel discussions optional rather than mandatory, and shrinking panel summary text to "three to five sentences."

For researchers, university research offices, and small businesses that have built their funding strategies around NSF, the question for the second half of 2026 is no longer whether NSF will fund their work. It is which categories of work the new NSF is structurally selecting for, which categories it is selecting against, and how a proposal needs to read differently to survive a review process that now operates with less external input, fewer reviewers, more program-officer discretion, and a panel summary that fits in a paragraph.

The Granted News brief on the FY2026 budget appropriation is at Granted News. This piece is the deeper analysis: the terminations data, the merit review changes, the leadership gap, and what it all means for proposal positioning in the back half of 2026.

What the Terminations Data Actually Says

The 1,752 terminations across April and May 2025 are not evenly distributed across the agency. Three patterns matter, and each one changes how prospective applicants should think about category selection for a 2026 proposal.

The first pattern is directorate concentration. The STEM Education directorate absorbed 839 of the 1,752 terminations, totaling $888 million — nearly half the dollar value of the entire termination wave. The Social, Behavioral and Economic Sciences directorate lost an additional 320 grants worth $91 million combined. Together, those two directorates accounted for the majority of terminated dollars and a clear majority of terminated grant counts. The Mathematical and Physical Sciences, Engineering, Computer and Information Science and Engineering, Biological Sciences, and Geosciences directorates were touched but not gutted — termination rates in those directorates ran in the single digits as a percentage of active awards.

The second pattern is topical concentration within directorates. Within STEM Education, terminated grants disproportionately targeted broadening-participation work, diversity-focused fellowship pipelines, and research on equity in STEM outcomes. Within SBE, terminated grants disproportionately targeted misinformation and disinformation research, social-network polarization studies, and research on online harms. The terminations were not random. They followed an explicit policy framework articulated in executive orders earlier in 2025 and were applied through agency-level review against active award abstracts.

The third pattern is mechanism. Terminations did not pass through scientific peer review or program-officer scientific judgment. They were administrative actions taken at the agency level, with limited per-grant scientific defense. For grantees, this means the institutional protection that NSF awardees have historically counted on — that an active award is "safe" once funded because revoking it would require a finding of scientific misconduct or material noncompliance — no longer holds in the same way. The expanded termination authority that OMB has proposed in its May 29 Uniform Grants overhaul (deep analysis here) would formalize this mechanism across all federal grantmaking agencies. The NSF terminations are the first large-scale demonstration of what a federal agency can do under the new framework before that framework is even finalized.

The Merit Review Overhaul: What Changed on December 15

NSF's peer review process was, for most of its history, structured around external scientific review as the load-bearing component of award selection. The standard pattern for a competitive program was three or more external reviewers per proposal, a panel meeting in which proposals were discussed and ranked, a panel summary that captured the consensus and dissents, and a program officer recommendation that was bound by the panel's input. Under that structure, the program officer's discretion was real but bounded, the external reviewer community was the primary check on quality, and the panel summary was the institutional record of the decision.

Effective December 15, 2025, that structure was significantly loosened. The minimum number of external reviewers dropped from three to two. One of the two minimum reviews can now be conducted internally by NSF staff, meaning that the floor for external scientific input on a proposal is now a single external reviewer. Panel discussions became optional rather than mandatory, meaning a program officer can now run a competitive program without convening a panel at all. Panel summaries were explicitly shrunk to "three to five sentences," meaning the institutional record of the decision is now a paragraph rather than a multi-page document. And program officers gained substantially more individual discretion in selecting from the funded pool.

NSF's stated rationale was operational: workforce reductions and proposal backlogs following the 2024 government shutdown had created a review bottleneck that the prior structure could not clear. The structural consequence, however, is that NSF in 2026 is a more program-officer-driven agency than it has been in decades. Program officers always had some discretion at the margins of the funded pool, but the new framework gives them substantially more latitude to weight proposals based on their own judgment of fit with directorate priorities.

For applicants, this changes the proposal positioning calculus in three concrete ways. First, the program officer relationship matters more than it used to. NSF has long advised contacting program officers before submission; in the new framework, that advice has shifted from helpful to essential. Six to eight weeks before submission is the right window — long enough for the program officer to actually engage with the question, short enough that the priorities they articulate will still be current at submission time.

Second, the proposal narrative needs to read well to a smaller number of reviewers who may have less specific subject-matter expertise. The historical NSF pattern of writing a proposal that would survive three deep technical reviewers and a panel discussion has shifted toward writing a proposal that needs to survive two reviewers and a paragraph-length summary. The opening pages of the proposal carry more weight; the broader-impacts narrative carries more weight; the program-officer-facing framing carries more weight.

Third, the alignment between proposal topic and current agency priorities matters more than it did under the prior structure. With less external review and shorter panel summaries, the program officer's interpretation of directorate priorities is more load-bearing in the decision. Physical sciences, engineering, computer science, and biological sciences faced significantly fewer terminations than education and social sciences, which suggests both lower funding risk and higher alignment with current administration priorities. For applicants who can credibly position their work in those directorates rather than in STEM Education or SBE, the structural risk profile of a 2026 NSF proposal is meaningfully different.

The Leadership Gap and What It Means for Predictability

Director Sethuraman Panchanathan resigned on April 24, 2025. NSF Chief of Staff Brian Stone has performed director duties on an acting basis since. As of June 2026, no permanent replacement has been named, and there is no public indication that the administration intends to nominate one in the near term.

The leadership gap matters for predictability. NSF's strategic direction has historically been set at the directorate-and-division level, with the director's office providing high-level coordination and external representation. An acting director can keep the agency running, but acting leadership cannot credibly commit the agency to multi-year priorities, cannot restructure directorate boundaries, and cannot defend the agency in budget negotiations the way a confirmed director can. The FY2026 appropriation outcome — Congress rejected a 55 percent cut and restored the agency to baseline — was a positive sign for the agency's institutional standing. But the FY2027 budget cycle will begin under the same acting leadership, and the political dynamics that produced the proposed 55 percent cut in FY2026 have not changed.

For applicants planning multi-year work, the leadership gap argues for staging proposals to the shortest credible timeline that still does scientifically meaningful work. Five-year proposals are not impossible, but they carry more residual policy risk than they did under stable leadership. Two-year and three-year proposals carry less of that risk and allow applicants to re-evaluate at renewal time against the actual policy environment.

The CAREER Deadline and Other Spring/Summer 2026 Windows

The most consequential near-term deadline for early-career researchers is the Faculty Early Career Development Award. CAREER's directorate-specific deadlines cluster around the third to fourth week of July 2026. The CAREER award has not been a primary target of the termination wave — early-career investigators in physical sciences, engineering, and computer science remain strongly aligned with current agency priorities — but the merit review changes apply in full to CAREER proposals. The CAREER proposal narrative needs to read well to two reviewers, the program officer relationship matters more than it did in prior cycles, and the broader-impacts narrative should be written with the new political environment in mind.

Other near-term windows include the preliminary proposal deadline for research incorporating social and behavioral processes in mathematical epidemiological models (June 19), the Growing Convergence Research target dates in September, and the International Research Experiences for Students windows in July and August.

The NSF SBIR/STTR program reopened on June 2 with a first full-proposal deadline of July 27 (deep analysis here). The NSF SBIR/STTR program is administratively separate from the directorate-level academic award programs and was not affected by the termination wave; the restart is, in many ways, the cleanest signal of where NSF intends to put new dollars in the back half of 2026.

Compliance Items That Will Bite in 2026

Three compliance items have changed under the new framework and will produce administrative rejections if missed. Research security training is now required before submission across all NSF programs — applicants who have not completed the training cannot submit. The equipment threshold rose from $5,000 to $10,000, meaning items that were previously classed as equipment may now need to be budgeted as supplies. Data sharing at publication is now required across all programs, formalizing what was previously a strong recommendation in many directorates. And AI tool use in proposal preparation must be explicitly documented, with NSF's updated misconduct guidance specifically addressing fabrication, falsification, and plagiarism involving AI-based tools.

The AI documentation requirement is the one most likely to be missed. The standard guidance is to include an explicit statement in the proposal about which AI tools were used, for what purpose, and how the outputs were verified. The conservative practice is to use AI tools sparingly in the proposal narrative itself, to verify all factual claims independently, and to err on the side of explicit disclosure rather than implicit understanding.

Positioning a 2026 Proposal Against the New NSF

The structural picture is consistent. NSF in 2026 has the same budget as NSF in 2024, a meaningfully altered program portfolio after the termination wave, a loosened merit review process that puts more weight on program officer judgment, an acting leadership team that cannot make credible multi-year commitments, and a compliance framework that has tightened on research security and AI disclosure.

For a researcher writing a proposal in the back half of 2026, the practical implications stack up:

Contact the program officer six to eight weeks before submission, with a specific scientific question and a one-page summary of the planned work. Use that conversation to test whether the work aligns with current directorate priorities, not historical solicitation language.

Stage the proposal to the shortest credible timeline that does scientifically meaningful work. Two- and three-year proposals carry less residual policy risk than four- and five-year proposals in the current environment.

Write the proposal narrative for two reviewers and a paragraph-length panel summary, not three reviewers and a multi-page summary. The opening pages, the broader-impacts narrative, and the program-officer-facing framing all carry more weight than they did under the prior structure.

Position the topic within directorates and divisions that faced lower termination rates. Physical sciences, engineering, computer science, and biological sciences are structurally lower-risk than STEM Education and Social, Behavioral and Economic Sciences in the current environment. Researchers whose work could credibly sit in either directorate should consider the lower-risk path.

Document AI tool use explicitly, complete research security training before the submission window opens, and budget equipment items at the new $10,000 threshold.

The NSF that exists in June 2026 is a fundable agency. It is not the same agency that existed eighteen months ago. Proposals that read as though the prior agency were still in place will be selected against by the new review process. Proposals that read as though their authors understand the new agency — its priorities, its review structure, its compliance environment, its leadership uncertainty — will be selected for.