The US grants landscape, 2026
What our catalogue shows about who publishes money in the United States, how much of it is stated rather than implied, and when it closes. Every figure is counted over the records this database holds on the day the page renders.
Measured 2026-09-07 · 1,775 records in scope · every source we read
Frozen 2026-09-07 · written once by the sweep and not recomputed since
Page computed 2026-09-07. The figures below are counted at read time over the inventory observed on that date. They are not a frozen snapshot: a later read may return different numbers, and the measurement date is what tells you which reading you are looking at.
Limits of this report
Every figure describes our catalogue, not the United States. Two feeds are read today; a state absent from the coverage table is absent from our corpus, not from American grantmaking, and no figure here should be read as a national total.
Groups below 10 records show their count and withhold their distribution. Fields the sources did not state are excluded from the denominators rather than counted as zero, which is why several counts here are smaller than the record total.
How a record is built, what verification means and what we refuse to infer are set out in our methodology, which this report does not restate.
What we hold
How many records, from which sources, as of the observation date.
Every record in our catalogue, federal and state. This describes our corpus, not the whole of US grantmaking -- the gap between the two is the subject of the limits section.
By level of government
Records carrying a level value, counted from grant_eligibility. A record with no level is absent rather than counted as zero. Measured 2026-09-07.
Based on 1,775 records
By source
Every record, grouped by the feed it was read from. Measured 2026-09-07.
Based on 1,775 records
By status
Every record, grouped by its current status. Forecasted and active are counted apart and never summed into one "opportunities" figure. Measured 2026-09-07.
Based on 1,775 records
What the money looks like
Ceilings, floors and programme pools across the catalogue.
Per-award ceilings only. A programme pool is the whole budget and a ceiling is what one applicant may receive; mixing them would produce a median no applicant could ever apply for.
Award ceilings, all records
Records stating a per-award maximum. Records stating none are excluded rather than counted as zero. Measured 2026-09-07.
Based on 786 records
Award ceilings, state
Records at the state level stating a per-award maximum. Measured 2026-09-07.
Based on 53 records
Award ceilings, federal
Records at the federal level stating a per-award maximum. Measured 2026-09-07.
Based on 733 records
When it closes
How deadlines fall across the calendar.
Deadlines by month
Records stating a deadline date, grouped by its UTC month across all years. Rolling and unstated deadlines are excluded and counted in the coverage section instead. Measured 2026-09-07.
Based on 995 records
Applicant types and industries
Which applicant types and industries the records classify to.
Applicant types
Distinct records per applicant class. The two source portals publish about thirty spellings between them and this site groups them into eight classes; the class page lists the wording each one covers. A record admitting several classes is counted under each, and counted once within a class. Measured 2026-09-07.
Based on 7,285 records
Industries
Distinct records per industry value, top ten. The classification is each source portal's own, not ours. Measured 2026-09-07.
Based on 2,128 records
Money actually paid out
What funders recorded paying out, as distinct from what they offered.
Award records are what a funder has published about money it has already given, which is a different and rarer thing than an open opportunity.
Published awards
Rows in funder_award, from the feeds that publish recipient lists. This is not a sample of US grantmaking; it is every award record we hold. Measured 2026-09-07.
Based on 4,716 records
How early we saw it
How much notice a forecasted opportunity actually gave.
A forecast lead time — the days between first seeing a record as forecasted and its opening — is Not stated, and the reason is worth naming precisely. It is not missing for want of records. It is missing for want of a column: a grant row carries its current status and no history of it, and extraction_run counts records rather than naming them, so nothing anywhere records the day a record stopped being forecasted.
Computing it needs a status-transition log. Until one exists this section will keep saying so rather than substituting the gap between first_seen_at and open_at, which measures when we first saw a record and not when it changed.