Who Actually Wins Federal Small Business Grants

1,966 of 155,387 new federal assistance awards went to small businesses in FY2025

This is the number that reframes federal grant advice. The federal assistance pipe is enormous and it is overwhelmingly not pointed at businesses: it funds state and local government programs, universities, hospitals, tribal governments, school districts and nonprofits. A small business competing for federal grant money is competing inside the 1.265% slice, not the whole. That is not a reason to skip federal programs — the 1,966 awards below are real and the money is substantial — but it is a reason to be precise about which programs actually award to companies, which is what the ranked table further down is for.

Two different numbers, and only one of them answers this question. The federal award database can be queried two ways, and they differ by a factor of 4.1. Awards signed in FY2025: 1,966 — the count of small businesses that actually won something that year, and the figure this page uses throughout. Awards active in FY2025, meaning any award that received a funding action including multi-year awards signed earlier: 8,020. The second number is the one a default query returns, and quoting it as the number of small businesses that won a federal grant would overstate the answer roughly 4-fold.

GrantCompass analysis of USASpending.gov award data, fiscal year 2025, retrieved August 29, 2026.

Federal awards to small businesses have fallen 34.4% since FY2021

The direction matters more than any single year's level, because it tells an applicant whether the thing they are preparing for is getting easier or harder. Fewer awards from a stable or growing applicant pool means more competition per award, and it means advice written from the FY2022 landscape describes a market that no longer exists.

What this does not say. A drop in awards is not by itself a drop in an applicant's chances, because the number of applicants is not public — the federal award record documents winners, never entrants. It also is not a statement about federal funding levels: award counts and dollar totals move independently, and a year with fewer, larger awards can carry more money than a year with more, smaller ones. Recent fiscal years continue to be revised for months after they close.

GrantCompass analysis of USASpending.gov award data, fiscal year 2025, retrieved August 29, 2026.

The median federal award to a small business was $324,984

Award sizes here span five orders of magnitude, so the distribution is the answer and the average is a trap. The chart below places each percentile on a logarithmic scale because a linear one would render everything below the 90th percentile as a single sliver against the largest award. Read it as: the middle half of winners banked between $274,954 and $1,028,046, and the headline-grabbing awards at the right-hand end are a different population that happens to share a dataset.

These are amounts actually obligated to named recipients, which makes them a different measure from the advertised ceilings published on program pages. A program advertising "up to $500,000" tells you the maximum one applicant could receive under its own terms; the figures here tell you what recipients received. The two diverge for structural reasons — tiered tracks, partial budgets, cost-share rules and appropriations running out mid-cycle — and the gap is why an advertised ceiling should be read as a ceiling. Our companion analysis of advertised award amounts measures the other side of that comparison.

GrantCompass analysis of USASpending.gov award data, fiscal year 2025, retrieved August 29, 2026.

Ten listings made 59% of all federal awards to small businesses

This is the most practically useful table on the page, because it separates federal programs that actually award money to companies from federal programs that mostly do not. A listing's small-business share — the fourth and fifth columns — is the ratio of awards that went to small-business recipients against every new award the listing made that year. A listing at 90% is a business-facing program. A listing at 3% is a program whose money mostly goes to governments and institutions, and where a small business is a marginal case however well the eligibility text reads.

How to read a CFDA number. A listing is an accounting unit, not always a program. Some listings map one-to-one onto a named program you can apply to; others are umbrellas covering everything one office funds, in which case the row describes the office's whole portfolio rather than a single opportunity. Of the 198 federal programs in the GrantCompass catalog, 142 carry a high-confidence listing assignment, and 20 of those map to a listing that no other program in our catalog shares. An award can cite more than one listing. Each award here is credited to its primary listing, so the awards column sums to the 1,966 national total, while the "all new awards" denominator counts awards citing the listing anywhere. Measured across the 20 largest listings the two conventions differ by 1.02%, so a small-business share here may understate by about that much.

GrantCompass analysis of USASpending.gov award data, fiscal year 2025, retrieved August 29, 2026.

California businesses won the most federal awards; 53 states and territories won at least one

Federal grant money is not distributed in proportion to where small businesses are. It follows where the research institutions, national laboratories, defense contractors and energy projects are, because those are the programs that fund companies. The practical reading is not that a business in a low-count state is disadvantaged by policy — most of these programs are nationally competitive with no geographic quota — but that the industries the federal government funds are themselves clustered.

How to read this table. The state is the recipient's registered business address, not necessarily where the funded work happens. A company headquartered in one state performing a project in another is counted at its headquarters. State-level grant programs — a far larger source of small business funding than the federal layer — are not in this dataset at all; those are covered on our state and federal program statistics page.

GrantCompass analysis of USASpending.gov award data, fiscal year 2025, retrieved August 29, 2026.

Department of Health and Human Services made 999 of the 1,966 awards

Agency concentration is the fastest way to work out where to spend your research time. A handful of departments account for most business-facing federal assistance, and each has a characteristic award size that tells you what kind of project it funds — research-scale, demonstration-scale, or capital-scale. An agency making many small awards is running a program with a wide funnel; one making few large awards is running a competition with a narrow one, whatever both call themselves.

Mean award size is used in this table rather than the median because it is what divides directly into the total, making the two columns readable against each other. Everywhere else on this page the median is the headline, for the reasons set out in the amounts section above.

GrantCompass analysis of USASpending.gov award data, fiscal year 2025, retrieved August 29, 2026.

Five conclusions the FY2025 award record supports

• Federal grants are a small-business rounding error by count. 1,966 of 155,387 new federal assistance awards, 1.265%, went to small businesses. The federal pipe is mostly aimed at governments and institutions, and any plan that starts at Grants.gov is starting in the narrowest part of the funnel.
• The number is falling, and faster than the pipe overall. Awards to small businesses fell 34.4% between FY2021 and FY2025, against 26.1% for all recipients over the same span. FY2022 saw 7,774 awards — 4.0 times FY2025's 1,966 — so advice drawn from that year describes a market that no longer exists.
• Quote the median, never the mean. The median award actually received was $324,984, with the middle half between $274,954 and $1,028,046. The $1,234,856 mean is 3.8 times the median because a handful of awards run into the tens of millions, and it describes no typical recipient.
• A handful of listings do most of the awarding. Ten of 118 assistance listings made 59% of all awards to small businesses. Identifying which listings actually fund companies is worth more than reading a hundred eligibility pages.
• Winning is not the same as applying. The federal award record documents recipients only. It contains no count of applicants, so no figure on this page — or on any page built from this data — describes an individual applicant's chance of being funded.

GrantCompass analysis of USASpending.gov award data, fiscal year 2025, retrieved August 29, 2026.

Methodology, sources and limitations

The source. USASpending.gov publishes recipient-level detail for every federal award: recipient name and unique entity identifier, obligated amount, awarding agency, assistance listing number and recipient location. It is a work of the US federal government, in the public domain, and requires no API key. Data is submitted by the awarding agencies and corrected retroactively, so figures for recent fiscal years continue to move after the year closes.

The filters. Award type codes 02, 03, 04 and 05 — block grants, formula grants, project grants and cooperative agreements. Small business means the award record carries the small-business flag the recipient self-certified in its SAM.gov registration; it is not an SBA size-standard determination made at the time of award. Fiscal year is the federal year, October 1 to September 30.

New awards versus active awards. The federal award API's default time filter returns every award that received any funding action in the window, including multi-year awards signed years earlier. For FY2025 that default returns 8,020 awards. This page instead counts awards whose base action was signed inside the fiscal year: 1,966. Those are different facts and only the second answers "how many small businesses won a federal grant that year". Both are published in the table above so the difference is visible rather than hidden.

The catalog join. Assistance listing numbers are matched to GrantCompass program records by backend/scripts/assign-cfda.py, which validates every number against the current SAM.gov Assistance Listings extract and records a confidence level and an evidence string per program. Of 198 federal programs in the catalog, 142 carry a high-confidence assignment; 49 carry none, most of them tax credits, certifications and contract vehicles that have no assistance listing by construction. No figure on this page depends on a low-confidence assignment.

Limitations, in order of how much they matter.

• The small-business flag is self-certified at SAM.gov registration and is neither verified nor re-checked at award time. It undercounts small firms that never set the flag and overcounts subsidiaries of large corporations that did.
• Federal assistance awards only. State, municipal, corporate and foundation grants — the large majority of the programs in the GrantCompass catalog — are not in this dataset at all and no share, rate or total here describes them.
• An award is not an applicant. USASpending records who won; it does not record how many applied, so nothing in this dataset describes an individual applicant's chance of being funded.
• Many CFDA numbers are umbrella listings covering a whole office's programs. Per-listing figures describe the listing, and program_specific marks where they cannot be attributed to a single catalog program.
• Federal award data is reported by agencies and is corrected retroactively; recent fiscal years continue to move for months after they close.

Related first-party data. Small business grant statistics and US small business funding statistics measure the programs themselves — what is offered, where and on what terms — across all funder tiers rather than the federal layer alone. Average small business grant amount covers advertised ceilings, the counterpart to the received amounts on this page.

GrantCompass analysis of USASpending.gov award data, fiscal year 2025, retrieved August 29, 2026.

How to cite this analysis

The underlying federal award data is in the public domain. This analysis of it is original first-party work and is free to cite, quote or reproduce with attribution and a link. Journalists, researchers and writers are welcome to use any number on this page.

Corrections and methodology questions are welcome. The scripts that produce every figure are named above, so any number here can be independently recomputed against the same public dataset.