Federal contracts reserved for small business were worth **$17.3B** in February through May 2024, and **$18.7B** in the same four months of 2026. The programme grew. Over the same two years every federal contract dollar grew nearly five times faster, and the share of the money reserved for small firms fell from **7.69% to 5.96%**. That much [our data study of the federal award record](/federal-contracting-shift) already says, and so does [the companion post](/blog/homeland-security-construction-spending-grew-35x). What neither says is *when* it happened, *where* it happened, or how much sharper it is in construction than in the federal government as a whole. All three are in the same snapshot, and each one changes what a small firm should do about it. ## The share survived year one. It broke in year two Three matched February–May windows, contract obligations, from USAspending: | | Feb–May 2024 | Feb–May 2025 | Feb–May 2026 | |---|---|---|---| | Reserved for small business | $17,276M | $15,262M | $18,708M | | All contract obligations | $224,606M | $212,988M | $313,671M | | **Reserved share** | **7.69%** | **7.17%** | **5.96%** | Read the middle column, because it is the one the study pages do not print. In year one the reserved programme shrank — $17.3B to $15.3B, down **11.7%**. But federal contracting shrank too, by 5.2%, so the reserved share slipped only from 7.69% to 7.17%. Half a point. A firm reading the award record in mid-2025 would have concluded that small business was having an ordinary bad year in an ordinary flat market. Year two is where it went. Reserved dollars rose **22.6%** — a real recovery, and the only reason the two-year figure is +8.3% rather than a decline. Everything else rose **47.3%**. The share lost another **1.21 points**, more than twice what it lost in the year the dollars actually fell. That inverts the intuitive version of the story. The reserved share did not erode through a year of cuts. It fell hardest in the year the money came back. Across both years the share is down **1.73 points**, or a little under a quarter of the share it held. Sized another way: had the 2024 share held through the expansion, February–May 2026 would have put **$24.1B** through small-business set-asides instead of $18.7B — a gap of **$5.4B in four months**. That is arithmetic on the published windows, not a forecast of anything. ## In construction, it halved Federal construction is the same movement, twice as sharp. Same windows, [its own study](/federal-contracting-shift/construction): | | Feb–May 2024 | Feb–May 2025 | Feb–May 2026 | |---|---|---|---| | Reserved for small business | $2,284M | $1,863M | $3,048M | | All construction obligations | $12,786M | $11,959M | $34,042M | | **Reserved share** | **17.9%** | **15.6%** | **9.0%** | Construction work reserved for small business rose **33.5%** over the two years — faster than the federal average, and faster than the reserved programme managed anywhere else. Federal construction as a whole rose **166%**. The reserved share did not decline; it halved. At the 2024 share, construction set-asides in the most recent window would have been **$6.1B rather than $3.0B**. Two things are true here at once, and a bidder needs both. Construction began with far more reserved than the rest of government — nearly 18 cents in every federal construction dollar against under 8 cents government-wide — so it had more to lose and lost more than half of it. And even after the halving, at 9.0% against 6.0%, construction is *still* the more reserved market. ## The growth landed where the reserved lane is narrowest — mostly This section changes measure, and the change is worth naming before the table rather than in a footnote. Everything above is matched February–May windows. The state figures below are federal **fiscal year to date — October through May**, because that is the window the [state profiles](/map) are built on, and FY2026 is an eight-month partial. The two measures are not interchangeable, and no sentence here mixes them. Reserved share of federal prime obligations, by where the work is performed, for states with at least $2B in the current window: | State | FY2024 | FY2026 | FY2026 reserved | |---|---|---|---| | [Alaska](/states/ak) | 22.8% | 22.8% | $0.5B of $2.2B | | [Maryland](/states/md) | 15.1% | 13.9% | $3.0B of $21.6B | | [Virginia](/states/va) | 9.3% | 7.4% | $4.8B of $64.2B | | [Texas](/states/tx) | 4.7% | **2.1%** | $1.6B of $76.5B | | [Arizona](/states/az) | 5.0% | **2.1%** | $0.5B of $22.6B | | [Connecticut](/states/ct) | 0.9% | 0.7% | $0.1B of $18.3B | Texas took the largest state increase in the dataset by a wide margin — $16.8B to $50.2B on the study's own windows, nearly three times the next state's gain. Arizona more than tripled. On the fiscal measure above, both reserve about **two cents in every federal dollar spent in them**, and both were near five cents two years ago. The pattern is not clean, and saying so is the point. **Virginia is the second-largest gainer at +$12.0B and reserves 7.4%**, above the national figure. Connecticut gained $4.5B at 0.7%. So the growth did not uniformly avoid states that reserve work — but the single largest destination of it is also the one that reserves least among the large states. We are not going to claim the second fact caused the first. The two figures sit on different windows, the award record carries no reason, and a state's reserved share is an outcome of what was bought there rather than a dial someone turned. What it means concretely is narrower and still useful: a firm re-pointing its pipeline at the growth is, in the two clearest cases, re-pointing it at the parts of the map where the reserved lane is thinnest. ## The climb of four years, given back in two Widen to seven fiscal years on that same October–May measure, whole of government: | FY | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 | |---|---|---|---|---|---|---|---| | Reserved share | 5.80% | 6.25% | 6.58% | 6.62% | **7.41%** | 6.73% | **5.89%** | The reserved share climbed for four years to a 7.41% peak in FY2024, then handed back almost the entire climb in two. FY2026 is the thinnest reading since FY2020 — and FY2020, at 5.80%, was thinner still. That last clause matters: this is a return to where the decade started, not an unprecedented low, and describing it as one would be the easiest available exaggeration. FY2026 is also a partial year and will move as it closes. ## The count of open work did not collapse Everything above is dollars. The number of things a small firm can actually bid is a different measure, and it points the other way. Both study pages carry a **live** count of open federal solicitations reserved for small business, read from our own corpus rather than from the snapshot — which is why there is no figure printed here. It would be stale before you read it. How that count is reported is worth knowing, because it is the reason to trust it. Most open notices never publish a set-aside field at all. So the page prints two ratios side by side — the share of *everything* open that is reserved, and the share of the notices that actually state a status — and reports how many stated nothing, rather than quietly folding them onto whichever side flatters the number. The two ends are far apart. Neither of them is small. ## What this changes for a bidder 1. **The reserved lane got narrower relative to the road, not shorter.** Reserved dollars are up 8% government-wide and 33% in construction. Nothing to bid has disappeared. What changed is that the unreserved market beside it grew four to five times faster, so the same proposal capacity spent the same way now captures a smaller share of a much larger market. 2. **A growth state is not automatically an opportunity state.** Texas absorbed more than a third of the national increase and reserves about 2% of its federal work. Follow the dollars there and you are competing unrestricted, against everyone, from the start. 3. **In construction, unrestricted competition is now most of the market.** Nine cents in the dollar is reserved where nearly eighteen were two years ago. Any capture plan built on the 2024 ratio is planning for a market that no longer exists. 4. **Read the newcomers as subcontract routes, not just competitors.** Five of the ten largest federal construction recipients did not appear in the top hundred two years ago. Firms that multiplied their backlog in eighteen months cannot self-perform all of it, and a subcontract on an unrestricted prime does not care about your size standard. Before any of that, the eligibility question is worth getting right, because it is checked late far more often than it is checked wrong: [what a set-aside actually excludes](/blog/what-a-set-aside-actually-excludes). ## What this does not show The measure is federal prime contract obligations, award types A through D. **Grants and loans are excluded**, so money reaching contractors through state agencies — including the 2021 infrastructure law — is invisible here. That is a fact about the measure, not about the spending. Every window stops at May 2026, because Defense and the Army Corps of Engineers withhold contract actions from the reporting system for 90 days. A more recent month reads as almost no defence activity; that is an embargo, not a decline. An earlier version of this study ignored it and published four inverted findings, one of which was this one — it reported that set-aside dollars *fell* 11% when they had risen 8%. The [methodology note](/federal-contracting-shift#method) sets out how the windows are bounded now. Obligations are not contracts signed: money added to an older award counts in the period it is obligated. Recent months revise upward as modifications post, so every figure here is a floor. Place of performance is where the source says the work happens, which for federal records skews toward the contracting office's own address. And this describes an award record. It does not explain one. Appropriations, procurement lead times and contract vehicles all sit between a policy and a number on this page, and none of them are visible in it. The full figures, charts and methodology are in the data study, in [all sectors](/federal-contracting-shift) and in [construction](/federal-contracting-shift/construction). What is reserved and open right now is on [the tender portal](/tenders), filterable by set-aside, with no account and no payment. ## Sources - [USAspending.gov](https://www.usaspending.gov) — every obligation figure here, queried through the public aggregation API. Prime contracts, award types A/B/C/D, matched February–May and October–May windows, snapshot 3 September 2026. - [FPDS: DoD data availability](https://www.fpds.gov/common/html/dodDataAvailability.html) — the 90-day hold on Defense and Army Corps contract data that bounds every window above.