NSF's Quiet Revolution: How DOGE Terminations and New Merit Review Rules Are Reshaping the Grant Game

Sethuraman Panchanathan resigned as director of the National Science Foundation on April 24, 2025. He reportedly did so after the White House ordered him to accept a 55% budget cut and eliminate half the agency's workforce. Nearly a year later, no permanent replacement has been named. The agency's chief of staff, Brian Stone, has been performing the director's duties since, steering an organization that has lost a third of its program officers, terminated over a thousand grants, and fundamentally changed how it decides which research to fund.

If you apply for NSF grants, everything you knew about the process 18 months ago is wrong.

The changes haven't received the attention they deserve, partly because they arrived in pieces — a policy memo here, a Dear Colleague Letter there — and partly because the far louder drama around DOGE's $49 billion grant termination campaign consumed most of the oxygen. But taken together, NSF's new merit review process (codified in NSF 26-200, effective December 15, 2025), combined with the agency's reduced workforce and political constraints, represents the most significant change to federal science funding procedures in decades.

What DOGE Did at NSF

DOGE personnel arrived at NSF headquarters in mid-April 2025. Within six weeks, the agency had terminated 1,752 grants with a combined value of approximately $1.4 billion.

The terminations were not random. Two directorates absorbed the vast majority of cuts. The STEM Education directorate (EDU) lost 839 grants worth $888 million — roughly two-thirds of its active portfolio. The Social, Behavioral, and Economic Sciences directorate (SBE) lost 320 grants worth $91 million. Together, these two directorates accounted for 66% of all terminations.

The targeting criteria, based on public reporting and court documents, focused on grants related to diversity, equity, and inclusion research and studies addressing misinformation and disinformation. One widely reported case involved DOGE flagging a $349,000 museum HVAC replacement grant as DEI-related after running the grant description through ChatGPT — a detail that emerged in subsequent litigation. The incident illustrated both the bluntness of the review methodology and the breadth of the net being cast.

For researchers outside the targeted categories, the terminations created second-order damage. Program officers who managed terminated portfolios were also managing non-terminated grants, and the organizational disruption — combined with workforce reductions that eliminated up to a third of program officers in some directorates — created processing backlogs that affected every applicant. Researchers with active grants reported delayed no-cost extensions, slower responses to budget modification requests, and increased uncertainty about renewal prospects.

Congress ultimately pushed back on the broader budget threat. The FY2026 appropriation, signed January 23, 2026, provided $8.75 billion for NSF — a 3.4% reduction from FY2024 levels, but a decisive rejection of the White House's proposed 55% cut to $3.9 billion. That congressional intervention preserved the agency's existence. It did not undo the institutional damage.

The New Merit Review: What Actually Changed

NSF 26-200, the policy document that governs how proposals are evaluated, rewrote rules that had been stable for over a decade. The changes are technical but consequential.

Minimum external reviews reduced from three to two. Previously, every NSF proposal had to receive at least three independent external reviews from qualified experts before a funding decision could be made. The new policy drops that minimum to two. In a system already strained by reviewer fatigue and declining participation rates, this means proposals may be decided on a thinner evidence base than at any point in recent memory.

One review may now be conducted internally. Under the old system, all reviews came from external experts — researchers at other institutions who had no stake in the funding decision. Now, one of the two minimum reviews can be conducted by NSF staff. This means a program officer could write one review, receive one external review, and have sufficient basis for a funding decision. The practical effect is a significant increase in individual program officer influence over outcomes.

Panel discussions are now optional. The traditional NSF review process involved external reviewers discussing proposals as a group — debating merits, surfacing concerns, and reaching a collective assessment. These panel discussions provided a check on individual bias and often surfaced issues that individual reviews missed. Making them optional rather than required means that some proposals will be evaluated without any collective deliberation.

Panel summaries limited to three to five sentences. When panels do convene, their written summaries — the documents that explain the collective assessment to applicants and program officers — are now capped at three to five sentences. Previous summaries routinely ran several paragraphs and provided substantive feedback that applicants used to strengthen resubmissions. The new limit effectively eliminates the feedback function.

Program officers gained substantially more discretion. The cumulative effect of these changes is clear: individual program officers now have more power to shape funding outcomes than at any point in NSF's modern history. With fewer required external inputs, optional deliberation, and minimal documentation requirements, the decision-making process has shifted from a consensus-driven peer review model to something closer to program-officer-driven selection.

NSF cited two justifications for the changes: the backlog of proposals awaiting review (roughly 20% more proposals were submitted in FY2025 than in FY2024, even as staffing declined) and workforce reductions that left some directorates without enough program officers to manage the traditional process.

What This Means for Your Next Proposal

The strategic implications are significant. Researchers who adapted their proposals to the old review system — writing for panel dynamics, anticipating group discussion, calibrating technical detail for diverse reviewer expertise — need to recalibrate.

Your program officer matters more than ever. Under the old system, a skeptical program officer could be overridden by enthusiastic panel reviews. Under the new system, the program officer is potentially both reviewer and decision-maker. This doesn't mean you should lobby — that remains inappropriate and counterproductive. It means you should understand your program officer's research priorities, read their Dear Colleague Letters carefully, and ensure your proposal connects to the programmatic themes they've identified. If your program has had staff turnover (many have), invest time in learning who the new program officer is and what they've funded.

Write for two readers, not five. When three to five panelists reviewed your proposal and discussed it as a group, you could count on at least one expert in your specific subfield catching the nuances. With potentially only two reviewers — one external and one internal — you need to write more accessibly. Assume your reviewers are intelligent scientists who may not be specialists in your exact area. Lead with the significance and broader impact. Reduce jargon. Make your methods comprehensible to a scientifically literate generalist.

The Broader Impacts criterion is now higher-stakes. NSF has always evaluated proposals on Intellectual Merit and Broader Impacts. In practice, Broader Impacts was often the weaker criterion — panels discussed it but rarely ranked it as decisively as technical merit. With program officers having more discretion and panels being optional, Broader Impacts becomes a tool that program officers can use to differentiate between technically similar proposals. Invest serious effort in this section. Connect your broader impacts to NSF's current strategic priorities (which you should be reading in the agency's strategic plan and recent policy documents, not guessing based on prior cycles).

Resubmission strategy has changed. The old process generated detailed panel feedback that applicants could use to systematically address weaknesses. Three-to-five-sentence summaries don't provide that roadmap. If your proposal is declined, you'll need to do more independent diagnosis of what went wrong. Request your individual reviews (you can still do this through Research.gov), and if possible, talk to your program officer about what would strengthen a resubmission. The formal feedback mechanism has shrunk; the informal one — direct communication with program officers — has become more important.

Budget and timeline adjustments. Several technical changes in the new guidelines create practical opportunities. The equipment threshold increased from $5,000 to $10,000, which simplifies budgeting for computing and lab equipment. Planning grant ceilings rose to $200,000, RAPID grants to $300,000, and EAGER grants to $400,000. If you're considering an exploratory or time-sensitive project, these higher ceilings may make NSF mechanisms more viable than they were previously.

Compliance is non-negotiable. The CHIPS and Science Act mandate requiring research security training for all senior and key personnel took effect October 10, 2025. Proposals that don't document compliance will be returned without review. Verify that your institution's training program meets NSF requirements, complete it before submitting, and include documentation. This is an administrative gate, not a judgment call — fail it and your science never gets evaluated.

CAREER Awards: The Exception and the Opportunity

NSF's CAREER award — the agency's most prestigious grant for early-career faculty — remains a high-functioning program with dedicated review panels. The late July 2026 deadlines (specific dates vary by directorate) give applicants time to adapt to the new review environment. CAREER proposals have always required strong integration of research and education, and in the current environment, proposals that demonstrate clear alignment with NSF's workforce development priorities and national competitiveness goals will have an advantage. The education plan is not a box to check — it's where you demonstrate the kind of broader impact that program officers are now empowered to weight heavily.

The FY2027 Threat: 50% and Counting

While Congress preserved NSF funding for FY2026, the FY2027 proposal asks for something even more dramatic: a cut exceeding 50%, reducing the budget to approximately $4 billion. The proposal would eliminate the entire Social, Behavioral, and Economic Sciences directorate — not just defund it, but abolish its research portfolio.

Congress rejected a similar proposal last year. But the combination of DOGE-driven terminations (which don't require congressional approval), workforce reductions (which have already occurred), and sustained political pressure against specific research areas means that NSF's effective capacity is already below its nominal budget. The agency funded approximately 8,800 new research grants in FY2025 — roughly 20% fewer than the 11,000 funded in FY2024. Even without the proposed 50% cut, the trajectory is downward.

Researchers should plan accordingly. Apply to NSF where your work fits, but don't stake your lab's survival on a single agency. The emergence of state-level research funds creates new options. Private foundations are expanding their programs. Industry partnerships, while carrying their own complications around intellectual property and publication, provide another funding stream.

The researchers who will thrive in this environment are the ones who treat funding as a portfolio problem — diversifying across federal, state, private, and industry sources rather than depending on any single stream. The ones who will struggle are those who continue to write proposals the way they did in 2023 and submit them to an agency that no longer works the way it did in 2023.

Granted helps researchers track funding opportunities across federal, state, and private sources — the kind of comprehensive view that's become essential when the traditional playbook no longer applies.