Two years ago the Department of Homeland Security obligated $431 million on construction in February through May. In the same four months of 2026 it obligated $15.3 billion.
That is not growth. That is a department acquiring an entirely new function, and it happened in under a year: in the same four months of 2025, DHS construction was still $250 million.
The firm that took the largest single share of it, Barnard Construction, does not appear anywhere in the hundred largest federal construction recipients of two years ago. In February through May 2026 it was the sixth-largest recipient of federal contract dollars of any kind, ahead of Huntington Ingalls and Northrop Grumman.
If you bid federal construction work and your target list still looks the way it did in 2024, it is pointed at the wrong agency, in the wrong state, and quite possibly at the wrong prime.
We pulled the whole picture apart in a data study built on the federal award record — agency by agency, firm by firm, state by state, with the methodology written out. This is the short version, aimed at the practical question: what does a bidder do differently now?
Nothing shrank. Everything got bigger, and construction got bigger fastest
Start by discarding the assumption underneath most conversations about federal spending since January 2025. Contract obligations did not fall. They rose 42% — $224.6B to $318.7B across matched February–May windows.
Federal construction rose faster still: $12.8B to $34.0B, a 2.7-fold increase.
And the growth is not one agency's story, even though one agency dominates the headline:
- Homeland Security, 3.5× across all sectors — $7.4B to $26.2B.
- Veterans Affairs +52%, Transportation +34%, Energy +20%.
- Defense +49%, and its construction spending alone up 70%, from $8.7B to $14.7B. The Pentagon did not lose this race; it was simply outrun from a much smaller base.
- USAID, −76%. The Small Business Administration's own contracting went net negative.
That Defense number deserves a flag, because we got it wrong the first time we published this. Defense and the Army Corps of Engineers withhold their contract data for 90 days for operational security, so any window running closer to the present than that shows almost no defence activity and reads as a collapse. It isn't one. Every window here now stops where defence has actually reported. The methodology note explains it in full.
Five of the ten largest construction contractors are new
This is the part that should change a teaming conversation.
Of the ten largest federal construction recipients in the most recent window, five do not appear anywhere in the hundred largest of two years ago:
| Recipient | Feb–May 2026 | Two years earlier |
|---|---|---|
| Barnard Construction | $5.47B | outside the top 100 |
| Fisher Sand & Gravel | $3.83B | outside the top 100 |
| Southwest Valley Constructors | $2.25B | outside the top 100 |
| Spencer Construction | $2.10B | outside the top 100 |
| Cochrane USA | $0.64B | outside the top 100 |
Others did not arrive so much as detonate. Thalle Construction went from $30M and 69th place to $1.74B and 5th — a 58-fold increase in two years.
It would be easy to read that turnover as an open market. The award record says otherwise. All five arrivals hold seats on one contract: the southwest-border construction contract that U.S. Customs and Border Protection awarded to eleven firms on 31 October 2025, two weeks after Homeland Security waived full-and-open competition, bid-protest and small-business rules for border-barrier work (90 FR 48281). CBP can send an order to a single holder. Barnard's $1.6B El Paso order went out that way, and the Government Accountability Office dismissed the protest against it in July 2026 (B-424472).
Each order is obligated almost in full the day it is signed, so a handful of signings carries a firm to the top of the table. Barnard's $5.47B comes almost entirely from six contract actions, and four border-wall orders make up $4.17B of it.
So the prime list did not open up. It turned over because one buyer funded one closed vehicle. The door that is open is the one below it: these firms multiplied their backlog in under a year and cannot self-perform all of it. Expect no help from the paperwork, though. The same waiver lifted the requirement that they file small-business subcontracting plans, so the subcontracts will not announce themselves.
Where the work physically is
Texas went from $16.8B to $50.7B. Arizona went from $3.7B to $11.5B — more than tripling. Virginia added $12.0B, Connecticut $8.3B, California $6.9B, Florida $5.4B, Washington $5.0B.
What is striking is the other side of the ledger: there barely is one. The largest decline of any state is Wisconsin at −$1.3B. The District of Columbia is down $0.4B. When the total rises this much, the question stops being who lost and becomes who captured the growth.
One caveat that matters if you plan around this: a federal award is recorded where the source says the work happens, which is not always where it happens. The bias is real and we describe it on each state profile. Directionally, though, a $34B move into Texas is not a data artefact.
Small business: the doors held, the share did not
Here is the number nobody quotes correctly, including us in an earlier version of this post.
Small-business set-aside obligations went up — $17.3B to $18.7B, a rise of 8%. Anyone telling you the set-aside pipeline collapsed in absolute terms is wrong.
But total obligations rose 42% over the same window. So as a share of all contract dollars, set-asides fell from 7.69% to 5.87%.
That is the real squeeze, and it is arguably worse than a headline decline would have been. The small-business programme did not shrink — it simply did not participate in the expansion. Nearly a quarter of the share it held two years ago is gone, in a period when the money on the table grew by more than ninety billion dollars.
Meanwhile the opportunity count has not collapsed. Of the federal solicitations open right now, thousands are explicitly reserved for small business — total small-business set-asides, SDVOSB, WOSB, HUBZone, 8(a) and tribal programmes. Among the notices that publish a set-aside status at all, the clear majority are reserved.
So the doors are still there. What has thinned is the money behind each one relative to everything else on the table. Bid selection matters more than it did: chasing the same number of set-asides while the unreserved market grows 42% is how a small firm burns a year of proposal capacity for a shrinking slice.
What to actually do
- Re-point the target list at the growth. DHS above all, then VA, Transportation and Energy. And do not write off DoD — its construction spending grew 70%, which the first version of this analysis missed entirely.
- Look at the newcomers as teaming partners, not competitors. Their prime seats are closed, but firms that went from near zero to billions in under a year have a capacity problem, and capacity problems are subcontracts. CBP's next construction vehicle is already signed too: a ten-firm design-build contract from May 2026, for work under the same law. Its holders are the next teaming list.
- Follow the work geographically. Texas and Arizona are where the construction dollars landed. Registration, bonding and labour in those states is a concrete, boring, decisive advantage.
- Get more selective inside the set-aside pipeline. The reserved work is still there and still growing; it is growing far slower than everything around it. Qualification is no longer the binding constraint — choosing well is.
The full analysis, with every figure, chart and caveat, is in the data study. Every number in it comes from USAspending.gov and can be checked against the same source: the study prints the query, and publishes the full snapshot as JSON.
Every open notice Offra tracks is free to search on a free account, and if you would rather be told when matching work opens than go looking for it, the account saves the search and emails you when something new matches it.
Sources
- USAspending.gov: every obligation figure here, queried through the public aggregation API. Prime contracts, award types A/B/C/D, matched February–May windows, snapshot 15 October 2026. Contract-level figures (Barnard's orders, the CBP contract's eleven seats) read from its award records on 20 October 2026.
- DHS waiver, 90 FR 48281: one of the nine sector determinations of 15 October 2025 that waived competition, bid-protest and small-business rules for border-barrier contracting.
- GAO, SLS Federal Services, B-424472: the July 2026 decision on Barnard's El Paso order, which sets out how the CBP contract was awarded and how its orders are issued.
- FPDS: DoD data availability: the 90-day hold on Defense and Army Corps contract data that bounds every window.


