Math Got the Most Awards of Anything NSF Released in August — 600 of Them. It Also Lost the First Tuesday in June, and That Is the Part Departments Should Worry About.

Count the awards, not the dollars, and the mathematical sciences come out on top of NSF's entire August release.

NSF 26-520, MPS Mathematical Sciences Research Programs, carries $200 million against an expected 400 to 600 awards. No other solicitation in the twelve-NOFO package NSF issued on August 17, 2026 comes close on volume. Biology has more money — $280 million — but expects 250 to 300 awards. Physics runs roughly 300 awards on $180 million. Astronomy funds about 140. Math is projected to make up to twice as many awards as any of them.

That is the correct shape for the discipline. Mathematics runs on small, single-investigator grants: a summer month of salary, a graduate student, some travel. An implied average of $333,000 to $500,000 in total award value, spread over three to five years, is roughly what a working DMS grant has looked like for years. Nothing about the money is a surprise.

The structural change underneath it is a different story.

Fourteen programs into one

NSF 26-520 consolidates most mathematical sciences funding at NSF into a single solicitation covering fourteen programs:

Disciplinary programs (11): Algebra and Number Theory · Analysis · Applied Mathematics · Combinatorics · Computational Mathematics · Foundations · Geometric Analysis · Mathematical Biology · Probability · Statistics · Topology

Structural programs (3): Mathematical Sciences Research Infrastructure · Mathematical Sciences Focused Research Groups · Mathematical Sciences Research Training Groups

Proposals are accepted anytime, with target dates that vary by program. There are no restrictions on the number of proposals per PI, co-PI, or organization — math joins engineering, physics, astronomy, geosciences and biology in the uncapped group, in contrast to chemistry's two-per-year quota and materials research's one-per-fiscal-year limit. Voluntary committed cost sharing is prohibited. Eligible applicants are U.S. institutions of higher education, non-profit non-academic organizations directly associated with research or education, and federally recognized Tribal Nations.

For an individual mathematician writing a standard grant, this is straightforwardly good news. Uncapped submission, no deadline, no cost share, no letter of intent. The friction is lower than it has been in a long time.

For a department, the last two items on the program list change something more fundamental.

What RTG and FRG actually are

Research Training Groups and Focused Research Groups are not single-investigator grants that happen to be larger. They are a different species of proposal, and they have run as separate, deadline-driven competitions for two decades.

RTG exists to widen the pipeline into the mathematical sciences — to increase the number of well-prepared U.S. citizens, nationals and permanent residents entering the field. A recent RTG competition ran a maximum award of $500,000 per year, made three to ten awards against roughly $10 million, and had a full-proposal deadline of the first Tuesday in June, annually.

FRG funds concentrated multi-investigator attacks on specific problems, at $150,000 to $500,000 per year for up to three years, with DMS expecting to fund approximately seven to ten awards annually and announcing them in April.

Both are department-scale efforts. An RTG proposal is not written by one person over a weekend. It requires a coalition of faculty across subfields, a coherent training design spanning undergraduates through postdocs, institutional commitments, mentoring plans, and a chair willing to spend political capital lining people up. In practice, it takes months.

That coordination happened because there was a date.

The forcing-function problem

Here is the asymmetry at the center of this solicitation, and it is genuinely different from the deadline story everywhere else in the August package.

For a single-investigator proposal, a deadline is mostly a cost. It generates proposals written because the date arrived rather than because the idea was ready. When NSF removed deadlines from four Earth Sciences programs in 2015, submissions fell from 804 in 2014 to 327 over the following eleven months — a 59 percent drop — and the field's read was that what disappeared was largely the marginal, defensive, may-as-well proposal. Removing the date filtered for conviction.

For a coordinated department-wide proposal, a deadline is not a cost. It is the coordination mechanism. "The RTG deadline is the first Tuesday in June" is the sentence that gets eight faculty into a room in February, gets the chair to commit a course release in March, and gets the letters written in May. It is the shared commitment device that makes a collective effort survive contact with everyone's individual schedule.

Delete the date and nothing forces the February meeting. The proposal does not get rejected; it gets deferred — to next semester, to after the search, to when the co-organizer is back from sabbatical. Anyone who has tried to sustain a voluntary multi-person effort without a fixed deliverable date knows exactly how this ends.

This is the specific risk NSF 26-520 creates for the training pipeline, and it is the opposite of the risk everyone has been discussing. The concern about rolling submission has been a flood of proposals. In the RTG and FRG lane, the plausible failure mode is a drought — not because departments stopped wanting the awards, but because the thing that made them start writing in February no longer exists.

The second-order effect: competing against everything else

There is a resource question stacked on top of the coordination question.

Under the old structure, RTG competed against RTG. Roughly ten awards were made against a dedicated pot of about $10 million, judged by a panel assembled to evaluate training design. FRG competed against FRG, on its own timetable, with its own review culture.

Consolidation into a single $200 million solicitation puts those structural awards inside the same envelope as 400-plus single-investigator grants. NSF has not published a per-program breakdown of the $200 million, which is itself the important fact: the allocation between individual research and training infrastructure is now an internal program decision rather than a published competition line.

That is not necessarily bad — DMS program officers have latitude to fund a strong training proposal whenever it arrives rather than only in the June cycle, and a genuinely excellent RTG no longer has to beat nine other RTGs in the same three-week window. But it does mean the number of RTG awards per year is no longer something you can look up before deciding whether to spend a semester writing one. That uncertainty is a real cost to department planning, and it should be the first question a chair asks a program officer.

What to do before the fall

Set your own date. If your department has been contemplating an RTG or FRG, pick a submission target now, put it in writing, and treat it exactly as you would have treated the first Tuesday in June. The single most valuable thing a chair can do in the next month is restore the forcing function NSF just removed. Rolling submission gives you the freedom to choose the date; it does not give you permission to not have one.

Ask the program officer for the number. Before committing faculty months, ask DMS directly how many training and focused-research awards it expects to make in the coming year and roughly what share of the $200 million is pointed at them. This is a fair question, program officers answer it, and the answer materially changes whether the effort is worth it.

Understand that "accepted anytime" is not "reviewed anytime." NSF 26-520 preserves target dates that vary by program. A proposal that misses its program's target date is not declined — it waits for the next review cycle. For a PI with a student graduating in spring, a deferred review and a decline are functionally identical outcomes. Find the target date for your specific program and treat it as the operative deadline.

Use the absence of a cap deliberately, not reflexively. Math has no per-PI limit, which means an individual can hold a proposal in analysis and another in probability simultaneously. The Earth Sciences evidence suggests most people will submit less, not more. The advantage in a no-deadline, no-cap division goes to whoever maintains a steady pipeline — one in review, one in revision, one in drafting — rather than clearing the desk each fall.

Pick the program before you write, and confirm it. Fourteen programs now sit behind one solicitation number. The routing decision — does this go to Foundations or Combinatorics, Applied Mathematics or Computational Mathematics — determines which reviewers see it, and it is no longer visible from the cover page of a program-specific solicitation. A short conversation resolves in fifteen minutes what the document cannot resolve at all.

Do not assume the old solicitation documents still govern. RTG's award ceiling, FRG's three-year limit, the April announcement date: these were terms of superseded solicitations. Verify every parameter against NSF 26-520 itself before building a budget around remembered numbers.

Mathematics came out of the August 17 release with more funded projects than any other discipline and fewer restrictions than most. For the individual investigator, that is close to an unambiguous win. For the departments that build the field's next generation, the win is conditional on doing voluntarily what a date on a calendar used to do for free — and the departments that recognize that in the next few weeks will be the ones still writing training proposals a year from now. Working out which of fourteen consolidated programs actually fits a project, and what each one now costs to enter, is exactly the kind of triage Granted is built to shorten.