NSF's Biggest Computing Award Now Comes With a Kill Switch at Year Five. Expeditions in Computing Preliminary Proposals Are Due December 2.

Buried in the 12 notices of funding opportunity NSF released on August 17 — the $1.5 billion foundational research package — is a single line item that most coverage reduced to five words in a bulleted list: CISE Expeditions in Computing.

It deserves more than a bullet. NSF 26-525 is the new solicitation for the largest single award the Computer and Information Science and Engineering directorate makes: up to $15 million over seven years, with 2 to 4 awards expected against a total of $60 million per competition. The preliminary proposal deadline is December 2, 2026. Full proposals, by invitation, are due July 27, 2027.

Compare that to the previous solicitation, NSF 20-544, and the top-line numbers are identical — $15 million, seven years, 2 to 4 awards, $60 million. Nothing changed. Which is exactly why the changes that did happen are easy to miss, and why they matter more than the dollar figure.

Three phases, one gate

The old Expeditions budget was, functionally, a large grant. The new one is a staged investment with an explicit decision point in the middle.

• Phase 1 — Acceleration: up to $3 million across Years 1–2
• Phase 2 — Steady State: up to $3 million per year for Years 3–5 ($9 million total)
• Phase 3 — Wrap-Up: up to $3 million across Years 6–7

Read those numbers carefully. The ramp is real: $1.5 million per year in the first two years, doubling to $3 million per year in the middle three, then falling back to $1.5 million per year at the end. An Expeditions project is not funded flat. It is funded like a program with a build phase, a peak, and a landing.

And Phase 3 is not guaranteed. The solicitation states that each project will be merit reviewed in two site visits, conducted in Year 2 and Year 5, and that "the site visit during Year 5 will assess the project progress and determine if the project warrants support for Years 6 and 7."

That is a $3 million contingency sitting at the far end of a seven-year commitment, adjudicated by a site visit five years after the money starts. Institutions that treat the full $15 million as bankable when they build cost-sharing arrangements, hire staff scientists on seven-year horizons, or promise facilities buildout are underwriting a risk the solicitation explicitly declines to underwrite.

The InTrans supplement is gone

The second change is stated flatly in the solicitation's summary of revisions: "The Innovation Transition (InTrans) funding opportunity for Expeditions removed."

InTrans was the mechanism by which Expeditions teams could pursue additional funding to move research outputs toward practical deployment — the translational tail on a basic-research program. Removing it does not eliminate translation pathways for CISE work; NSF's Technology, Innovation and Partnerships directorate now runs several. But it does mean an Expeditions award no longer carries a built-in on-ramp to transition funding, and teams that were counting on InTrans as the Year 6–7 bridge now need to plan for TIP-side or industry-side follow-on from the beginning.

The combination is notable: the tail of the award got both smaller in structure (wrap-up phase, reduced annual rate) and more conditional (Year 5 gate), and the supplement that used to extend beyond it was withdrawn. Expeditions has become more sharply a research program and less a research-to-practice pipeline.

The calendar moved, and that costs you a year

NSF 20-544 ran preliminary proposals in June with full proposals the following February. NSF 26-525 runs preliminary proposals in December with full proposals the following July, biennially thereafter — first Wednesday in December, fourth Tuesday in July.

For anyone assembling a team, that reshuffles the planning year. A June prelim let faculty use the spring semester to build consortia and the summer to write. A December prelim means the consortium must be assembled during the fall term, with the 11-page preliminary project description written in the run-up to finals and Thanksgiving. Then the invited teams write full proposals across spring and into summer 2027, with awards realistically landing in late 2027 or 2028.

If your research vision has a two-year clock on it — a hardware generation, a policy window, a competitive race — Expeditions is not the instrument. The gap between "I have the idea" and "I have the money" is roughly 20 months at best.

What the preliminary proposal actually has to do

Preliminary proposals are required, and the project description is capped at 11 pages. Three things must land in that space:

An overarching vision where the whole exceeds the parts. This is the sentence NSF has used to distinguish Expeditions from large collaborative grants for over a decade, and it remains the single most common failure mode. Panels routinely reject prelims that read as four excellent single-PI projects sharing a cover page. The test is whether removing any one thrust collapses the vision. If the project survives the removal, it was a collection, not an Expedition.

A leadership structure with a named project coordinator. Not a vague management plan — an identified coordinator and a described decision structure. At $15 million across seven years and typically five-plus institutions, NSF is underwriting organizational capacity as much as science, and the Year 2 site visit will test it.

Shared experimental systems or facilities, where applicable. Expeditions with a common testbed, dataset, or fabrication capability have historically fared well because the shared artifact is the physical proof that the collaboration is real.

Add a one-page budget summary covering the full seven-year period. Build it against the phase structure, not as a flat annual line — a budget that spends $2.1 million a year for seven years signals you did not read the solicitation.

Eligibility and the one-proposal rule

Lead institutions must be U.S. institutions of higher education with undergraduate, master's, and doctoral programs in computer science fields. Subawardees may include two-year colleges, nonprofits, and other research entities. No cost sharing is required or permitted — do not offer it.

The binding constraint is personal: an individual may serve as PI or co-PI on no more than one preliminary and, if invited, one full Expeditions proposal per competition. There is no cap on proposals per institution. That asymmetry is strategic. Large departments can and do field multiple Expeditions teams, but a senior faculty member who is everyone's natural co-PI can only be on one. Consortium formation this fall will be, in practice, a competition for a small number of people who anchor credibility — and those commitments will be locked well before December.

Start the conversations that determine who anchors which team now, not in November.

Who this is for

Expeditions has funded projects like computing with living neurons, superconducting electronics design, computational decarbonization of infrastructure, learning-directed operating systems, and pervasive computational epidemiology. The pattern is consistent: a ten-year research agenda in an area that does not yet have a field around it, pursued by a coalition large enough to build the field while doing the work.

If your proposal fits comfortably inside an existing CISE core program, it is not an Expedition — and with those core programs now accepting proposals on a rolling basis under the new portfolio approach, the fallback path is more forgiving than it used to be. If it requires a decade and a coalition and would look reckless as a single-PI grant, this is the one federal instrument in computing sized for it.

Just build the plan around $12 million over five years, and treat the last $3 million as the reward for a good Year 5 site visit. Because that is precisely what the solicitation says it is.