NSF's Biology Postdoc Fellowship Just Deleted All Three of Its Research Tracks and Replaced Them With AI. You Get Two Lifetime Shots.

The National Science Foundation's Postdoctoral Research Fellowships in Biology program has run on the same architecture for most of a decade: several competitive areas, each with its own intellectual charter, each drawing from a different corner of the Directorate for Biological Sciences. Applicants picked a lane.

NSF 26-504, the solicitation whose first deadline fell on September 29, 2026, has one lane. All three competitive areas are gone, replaced by a single focus area: Intersection of Artificial Intelligence and Biological Sciences to Strengthen and Safeguard Biotechnology Innovations.

For a program whose eligibility window is measured in months and whose lifetime submission cap is two, that is not a thematic refresh. It is a redefinition of who is eligible to be an NSF biology postdoc at all.

What changed, line by line

The predecessor solicitation, NSF 24-593, offered three competitive areas:

1. Broadening Participation of Groups Underrepresented in Biology
2. Integrative Research Investigating the Rules of Life Governing Interactions Between Genomes, Environment and Phenotypes
3. Plant Genome Postdoctoral Research Fellowships, funded out of the Plant Genome Research Program in the Division of Integrative Organismal Systems

Roughly $14 million supported Areas 1 and 2; up to $3 million supported Area 3. About 60 fellowships per year. Each award paid $70,000 in annual stipend plus a $20,000 research and training allowance — $90,000 a year — across 36 continuous months, for a total of about $270,000.

NSF 26-504 reads differently on nearly every dimension:

The stipend is up 21 percent. The research allowance is up 25 percent. And the total value of a fellowship fell 18.5 percent, because the duration was cut by a third.

That trade is the quiet substance of the change. NSF is paying individual postdocs meaningfully better per year, supporting 25 percent more of them, and buying a third less time from each. In a field where a competitive independent research program frequently needs three years of postdoctoral runway to generate a publication record and a preliminary-data package, moving from 36 to 24 months shifts real risk onto the fellow. A 24-month fellowship that ends in the middle of a faculty job cycle is a 24-month fellowship that needs a bridge.

The focus area NSF announced — and the one it delivered

Here is the detail worth flagging for anyone trying to forecast where BIO is heading.

NSF 24-593 did not leave the future of Competitive Area 2 ambiguous. It stated that FY 2025 was the final year for the Rules of Life focus, and it named the successor: submissions at the next deadline would compete under "Biological research, understanding and solutions for a Resilient Planet."

That focus area never shipped. The slot it was announced for is now occupied by AI and biotechnology.

Pre-announcing a successor topic and then substituting a different one is not a procedural violation — solicitations supersede prior solicitations, and NSF reserves that right explicitly. But it is a useful data point about how much weight to place on forward-looking language in a BIO solicitation. If you were a 2024-vintage graduate student structuring a dissertation around resilience and planetary-scale biology because NSF told you where the money was going, the money went somewhere else.

The elimination of the broadening-participation area is the more legally conditioned change. That track was a demographically targeted competitive area, and a 2026 Department of Justice Office of Legal Counsel opinion concluded that several NSF programs structured around race-conscious participation criteria were unconstitutional — a sequence we traced in the OLC opinion that put $104 million of NSF STEM education funding in question. Reading the PRFB consolidation in isolation misses that context. Area 1 was unlikely to survive in its prior form regardless of what replaced it.

Area 3's disappearance has a third explanation entirely: the Plant Genome money came from IOS, and IOS no longer exists as a funding division. The BIO directorate dissolved DEB, IOS, and MCB into a single consolidated core solicitation, which we covered in NSF 26-517 BIO-Core. A program funded out of a dissolved division does not get a graceful retirement; it gets absorbed.

Three competitive areas, three distinct causes of death, one AI-shaped replacement.

What the new focus area is actually asking for

The solicitation frames the opportunity around a thesis: "The integration of Artificial Intelligence (AI) and Biological Research has the potential to pave the way for breakthroughs in biotechnology." NSF wants fellows who will "work seamlessly at the intersection of AI and biology."

Reviewers are instructed to assess alignment across three dimensions:

1. Using AI to analyze biological systems — the method half
2. Generating insights with technological potential — the biotechnology-translation half
3. Developing AI-literate biological researchers — the training half, which is really a judgment about the sponsoring scientist and host environment

The solicitation also points applicants toward leveraging existing biological data and infrastructure, naming Biofoundries and NEON specifically. That is a strong hint, and a cheap one to act on. A project that proposes new data collection as its primary activity is swimming against the stated intent; a project that proposes to extract previously inaccessible structure from an existing national-scale dataset is swimming with it. This is the same logic NSF applied in its $100 million dataset program, which we analyzed in NSF's $183 million bet on the data layer of AI science.

Note also the word Safeguard in the focus-area title. It is doing work. Biosecurity and biotechnology-risk framing is an explicit part of the charter, not decoration, and a project that engages it seriously has a differentiator that most AI-plus-biology proposals will leave on the table.

The eligibility math is the real constraint

This is where the consolidation becomes genuinely hard for individuals, and it has nothing to do with the science.

PRFB eligibility requires U.S. citizenship, national status, or permanent residency at submission. It caps you at 15 full-time months of work in positions requiring a doctorate. And it limits you to one submission per year and no more than two in a lifetime. The deadline is now a fixed annual event: September 29, 2026, then September 28, 2027, then the fourth Tuesday in September annually thereafter.

Stack those together. A new PhD has, in practice, one or two September deadlines inside the 15-month window before eligibility closes. There is no strategy available in which you wait out an unfavorable focus area. Under the old three-area structure, a candidate whose work did not fit Rules of Life could compete under Plant Genome or broadening participation. Under NSF 26-504, a candidate whose work is not credibly AI-adjacent has no area to compete in, and no later year in which a different area will open before they age out.

The practical consequence: the pool of eligible biologists who can win this fellowship shrank far more than the 75-award count suggests it grew. More awards, drawn from a narrower population, is a different program wearing the same solicitation name.

Preparing for September 28, 2027

The 2026 deadline has passed. The next one is twelve months out, which is the right amount of runway for this particular application because so much of it is not about you.

The sponsoring scientist statement is three pages and it is scored. Reviewers assess mentoring quality and host-institution suitability as named criteria. Every sponsoring scientist must also supply a biographical sketch, Current and Pending Support, and Collaborators and Affiliations information. If you have multiple sponsors, the statement is integrated rather than multiplied — three pages total, not three pages each. Choose a co-mentor who adds genuine AI methods depth rather than a second name, because the page budget punishes redundancy.

The project description is six pages with mandatory separate Intellectual Merit and Broader Impacts sections. NSF has made these distinct headed sections rather than themes to be woven through. Write them as the solicitation structures them; reviewers will look for the headings.

"Distinctness from doctoral research" is an explicit review criterion. This is the criterion most PRFB applicants underweight. You must include a one-page doctoral research abstract, and reviewers compare it directly to your proposal. A fellowship application that reads as the natural continuation of your dissertation loses on a stated criterion no matter how good the science is. If you are planning now for 2027, the pivot — toward AI methods, toward a new system, toward a different scale of data — is the thing to start building this year, not next August.

Two reference letters, maximum, three pages each. A cap, not a floor. Two excellent letters beat two excellent letters plus padding, and there is no third slot to fill.

For early-career researchers mapping which federal fellowship architecture still fits their work after a year of consolidations like this one, Granted tracks the live solicitations and the eligibility fine print that decides whether you have one shot, two, or none.