America's Seed Fund at NSF does not tell you what to build. It funds deep technology across nearly any sector — but you cannot submit a full proposal until a three-page Project Pitch has been invited.
Most SBIR agencies let you submit a full proposal whenever a solicitation is open. NSF does not.
You must first submit a Project Pitch — four short fields totalling roughly three pages — and receive an emailed invitation before you may submit a full Phase I proposal. No invitation, no proposal.
This reads as bureaucracy and is closer to a favour. Writing a full SBIR proposal costs weeks. The pitch costs an afternoon and gets you a scope decision from NSF program staff before you spend those weeks. If NSF declines, it tells you why the project is not appropriate for the program — which is a more direct answer than most agencies ever give, and more useful than most paid advice.
The constraint is that you may only have one pitch in play at a time and only two per year. Treat it as a real submission rather than a formality.
The most important thing to understand about NSF SBIR is what it is not. NSF states plainly that the programs do not solicit specific technologies and do not procure goods or services from startups. There are no government topics to answer.
You propose your own innovation. NSF funds high-risk, high-potential deep technology across nearly any sector, judged on whether the technical risk is real and whether resolving it opens a market.
That makes NSF the natural first stop for a company whose technology came out of its own research rather than out of a government requirement — and it makes NSF a poor fit for a company hoping to be told what the government wants to buy. For that, the DoD route is the opposite model.
Phase I is up to $305,000 over 6 to 18 months. Phase II is up to $1,250,000 over roughly 24 months.
Read those as envelopes, not research budgets. Both ceilings include direct costs, indirect costs, the small business fee, technical and business assistance funding, and — in Phase I — optional I-Corps participation. Applicants routinely build a budget around the headline number, then discover the fee and TABA come out of the same pot. Work backwards from the ceiling with every component included before deciding what research the award can actually buy.
NSF runs three submission windows a year. Published dates include July 27, 2026 and November 4, 2026, moving to the first Wednesday in November annually thereafter; March 4, 2027 and then the first Thursday in March annually; and July 7, 2027 and then the first Wednesday in July annually. Proposals are due 5 p.m. local time of the submitting organization.
An invitation from a Project Pitch stays valid for the next two submission deadlines. That is generous — roughly eight months of flexibility — but it does expire, and companies do let invitations lapse while waiting for a better moment.
Companies without a genuine technical risk. NSF funds R&D, not deployment. A proposal to build something that is understood to be buildable will not score.
Companies whose PI cannot commit. The PI’s primary employment — at least 51% — must be with the small business at the time of award. If your technical lead is a professor keeping their appointment, STTR is the vehicle, not SBIR.
Anyone who wanted to skip the pitch. There is no path directly to a full proposal, and asking about one wastes a cycle.
It is short by design — roughly three pages worth of text across four fields. NSF uses it to tell you early whether you are in scope, which saves you writing a full proposal for a program that was never going to fund you.
You may only have one Project Pitch in play at a time, and only two per year. If NSF declines, it tells you why the project is not appropriate for the program. That feedback is more useful than most paid consulting.
An invitation remains valid for the next two submission deadlines. NSF runs three windows a year, so an invitation gives you real scheduling flexibility — but not unlimited.
NSF funds high-risk R&D with commercial potential. The proposal must convince reviewers both that the technical risk is real and that resolving it opens a market.
The ceiling includes direct costs, indirect costs, the small business fee, technical and business assistance, and optional I-Corps participation. Applicants routinely forget the fee and TABA and end up with less research money than they planned.
