Construction Market Intelligence — September 4, 2026
July construction spending fell 3.8% y/y and input costs held at 7.1% — double the 3.5% rise in bid prices — as backlog slipped to 8.0 months and September Fed-hike odds jumped toward 60%.
Three releases landed in four days and they point the same way. The Census Bureau reported that construction spending fell to a $2,157.6 billion seasonally adjusted annual rate in July — 3.8% below a year ago — but the decline is not broad; it is one sector collapsing (manufacturing −21.2%) while another booms (office, which carries data centers, +16.9%). Strip data centers out and nonresidential spending is at its lowest since September 2023. Second, cost: the producer price index for inputs to new nonresidential construction is up 7.1% over twelve months — roughly double the 3.5% rise in contractors’ bid prices, a gap that is coming straight out of margin. Third, capital: a strong August jobs report pushed the two-year Treasury to its highest since January 2025 and lifted the implied odds of a September rate hike to roughly 60%. Underneath it all, contractor backlog fell to 8.0 months, the lowest since January. This is a margin-compression market, not a volume-recession market.
Six-signal dashboard
The signals this week are not offsetting — they compound: falling volume, elevated unit cost, firmer labor and a cost of capital drifting the wrong way.
Market signals
The uncomfortable feature of this week’s evidence is that the signals do not offset — they compound. Demand is contracting in every category except the two most contractors cannot bid. Input costs are running at roughly twice the rate contractors can push into price. Craft labor is simultaneously scarcer and more expensive, the industry adding jobs for a sixth straight month even as spending falls. And the capital markets, expected for two years to ease and relieve feasibility pressure, are now pricing the opposite. Falling volume, rising unit cost, tightening credit is a margin-compression environment, not a volume-recession one. Firms will not fail this cycle for lack of work; they will fail on jobs already signed, priced against escalation assumptions that no longer hold and financed against rates that have moved. The counter-signal worth holding: the labor market surprised sharply to the upside, and construction employment is growing at its fastest year-over-year pace since February 2025 — real weakness is not yet visible in payrolls.
Cost watch
The latest PPI reading is July; August is due mid-September. Inputs to new nonresidential construction are up 7.1% over twelve months — the same 7.1% recorded in June, a plateau at an elevated level rather than an acceleration, but a plateau well above what most 2026 escalation clauses were written to absorb. Against that, contractor bid prices for new nonresidential buildings rose just 3.5%. The spread between what contractors buy and what they sell has held near 360 basis points for two straight months, and it is coming out of margin. Diesel is up 44.2% year over year — the single most volatile line in any estimate (retail diesel was $5.58/gal on Aug 21 versus $5.40 on Mar 30). Lumber and plywood are up 9.9% (most since March 2022), paving mixtures 6.6% (a three-year high), and construction plastics 5.0%. From the June reading, tariffed metals remain extreme: aluminum mill shapes +52.4%, copper and brass +26.0%, steel mill products +16.9%, all carrying Section 232 duties up to 50%.
Two changes are warranted now. Reprice escalation language written against a 3–4% assumption, or convert it to indexed allowances tied to a published series — the AGC/BLS input index is the defensible reference. And isolate fuel as a discrete risk line rather than burying it in general conditions; at a 44% year-over-year move, diesel is its own exposure on any earthwork-, paving- or haul-intensive scope. For owners, the 7.1%-versus-3.5% spread is the number to watch: today’s bids are being subsidized by contractor balance sheets, and that subsidy is finite — expect the next two quarters of bids to close the gap.
Labor & capacity
August payrolls, released September 4, showed construction adding 22,000 jobs — a sixth consecutive monthly gain, with nonresidential specialty trades up 8,000 against a 12-month average of 6,000. The composition is telling: on ENR’s reading of the BLS detail, the segment shedding jobs is nonresidential building (roughly −1,800), precisely the segment the spending data shows contracting, while residential building added jobs. Construction average hourly earnings reached $41.46 in July (the latest construction-specific month), up 4.4% year over year, against +3.1% for all-private employees in August — craft pay is outrunning the broad market by more than a point. AGC’s September 4 release headlines a record-low construction unemployment rate of 3.1% (not seasonally adjusted; reported as AGC characterizes it). JOLTS for July showed construction openings up 28,000 even as backlog fell — demand is rotating between segments, not disappearing, against a genuine trades skills constraint.
Labor is the one input where contractors currently hold no leverage and are unlikely to gain any. Hiring is concentrated in nonresidential specialty trades chasing data-center and power work, so electrical, mechanical and controls crews are being bid away from conventional commercial and institutional projects at a premium. Plan 4–5% craft wage escalation into 2027 budgets, not 3%; and if you bid against a data-center-exposed subcontractor base in the same market, verify crew availability by name and by date — a sub carrying 11.4 months of backlog is not the same risk as one carrying 7.5.
Procurement watch
Lead-time evidence here is drawn from supplier, contractor and consultant reporting rather than an official series — treat the ranges as directional and verify per manufacturer and per project. Distribution transformers have improved to roughly 30 weeks (from a 2023 peak above 100) — the one genuine easing. Everything upstream of them has not: switchgear averages around 44 weeks but standard switchboards run near 52, power circuit-breaker switchboards past 84, and 15 kV medium-voltage switchgear 52–80 weeks (data-center configurations two to three years). Power transformers sit near 128 weeks, generator step-up units near 144. The downstream effect is measurable: of roughly 12 GW of U.S. data-center capacity slated for 2026, only about 5 GW is under active construction, with transformer, switchgear and battery shortages among the leading causes.
On any project with meaningful electrical scope, the governing date is no longer substantial completion — it is the switchgear release-to-fabrication date. Release long-lead electrical procurement before design is complete, using performance specs and owner-direct purchase where the contract permits, and treat the change-order exposure as cheaper than the delay. Write liquidated damages and milestones against energization, not occupancy. And if you are underwriting a 2027 delivery, a 128-week transformer means the order should already be placed — if it is not, the pro forma delivery date is not real.
Capital & development
The federal funds range has held at 3.50–3.75% since September 2025, but this morning’s payroll beat moved the two-year Treasury to 4.390%, its highest since January 2025, and lifted the implied odds of a 25-basis-point hike at the September 15–16 meeting to roughly 60%, from about 52% beforehand. The ten-year sits at 4.782% after touching a three-year high early in the week on rising energy prices and hawkish remarks from the Fed chair. The 2s10s spread is just 39 basis points — a curve that narrow prices in very little easing over two years. The committee is not unanimous: Governor Waller has said he would support a hold if price pressures keep easing, and that his vote turns heavily on next week’s August CPI. Prime stands at 6.75%, SOFR at 3.66%. Development-side discipline is tightening too — major lenders are now pricing municipal-opposition and zoning risk into data-center underwriting, a discipline absent two years ago.
Do not translate a 5-basis-point move in the two-year into a construction-loan repricing — that is not how transmission works. What has changed is direction. For two years the base case in every feasibility model was that financing costs would fall and rescue marginal deals; market pricing no longer supports that. Re-run any 2027 pro forma flat, and stress it 50 basis points higher. Both index rates that matter — prime at 6.75%, SOFR at 3.66% — now carry upside rather than downside risk, which puts yield-on-cost, not exit cap rate, at the center of underwriting.
Backlog & pipeline
The ABC Construction Backlog Indicator fell to 8.0 months in July, down 0.8 from June and the lowest since January; every region declined month over month, with the South highest at 10.1. The dispersion is the story: the 12% of contractors on data-center work report 11.4 months, the other 88% report 7.5; firms above $100M hold 12.1 months, those under $30M hold 7.0, and the $30–50M band fell to its lowest since March 2020. Forward planning is more constructive — the Dodge Momentum Index rose 6.9% in July to 291.7 (+11.7% YoY), institutional planning up 13.1% for the month — but excluding data centers, Dodge’s commercial segment would be down 16.2% year over year. Design-side evidence is weaker still: the AIA/Deltek Architecture Billings Index registered 46.6 in July, below the 50 breakeven, with contracts at 47.2 against inquiries at 52.6 — clients wanting to pursue projects they are not prepared or able to finance.
A national average conceals a bifurcated pipeline entirely. A mid-market GC in a market without an AI campus is not carrying the headline 8.0 months — it is closer to 7.5, nearly four months less work than the data-center-exposed contractor it competes against for crews, while absorbing input costs rising 7.1%. Know which side of that line your book sits on; it explains more about your 2027 than any national indicator will.
What changed this week
- August payrolls beat consensus by more than 100,000; June–July revised up 55,000. Implied September-hike odds moved from ~52% to ~60% in a session and the two-year Treasury hit its highest since January 2025. The direction of the next Fed move has flipped — the Fed has decided nothing.
- Construction spending fell to $2,157.6B SAAR, −3.8% y/y (July, released Sept 1). Year to date is −3.5%; excluding data centers, nonresidential is at its lowest since September 2023 — the underlying market is weaker than the aggregate.
- Construction added 22,000 jobs in August — a sixth straight gain. Labor is not confirming the spending weakness; craft wages at +4.4% y/y (July) vs +3.1% all-private (August) mean no labor-cost relief in 2027 budgets.
- JOLTS showed construction openings up 28,000 in July. Openings rising while backlog falls is consistent with demand rotating between segments, not disappearing — and with a real trades skills constraint that data-center work is absorbing.
- Federal surface-transportation authorization expires in weeks (Sept 30). The House approved the five-year, $580B BUILD America 250 Act (H.R. 8870) 62–2 on May 22, but the Senate has not released a proposal — highway and heavy-civil contractors face real reauthorization uncertainty.
What we’re watching
- August CPI, due next week. Governor Waller has said his September vote turns heavily on it. With the hike near 60% and the committee divided, this is the single largest near-term input to 2027 financing costs.
- FOMC, September 15–16. A hike lifts prime to 7.00% and reprices prime-indexed construction facilities within a billing cycle; SOFR-indexed facilities reset on their own schedule.
- ABC Construction Backlog Indicator, August reading (due within days). A second straight decline would confirm a trend and be the clearest sign yet that bidding conditions are shifting toward owners outside the data-center segment.
- Surface-transportation reauthorization, Sept 30. Watch for a short-term extension versus a lapse; CBO projects the Highway Trust Fund’s highway account approaches zero in FY2028 regardless.
- August construction-input PPI, due mid-September. Two months at +7.1% established a plateau; a third confirms it, and a move above 8% would make most 2026-signed escalation assumptions indefensible.
Executive takeaway
This week did not deliver a shock; it delivered confirmation, which is more useful and more dangerous. The market entering the fourth quarter of 2026 is not one market — it is a data-center and power economy running at capacity against hard equipment constraints, and a conventional building economy contracting under input costs it cannot pass through. A contractor’s position on that divide explains more about its 2027 than any national indicator: 11.4 months of backlog versus 7.5 is not a rounding difference, it is a different business. For everyone on the wrong side of the line, three exposures now stack: an input index rising 7.1% against bid prices rising 3.5%; craft wages rising 4.4% with no relief in sight; and a cost of capital whose expected direction has just reversed. None is catastrophic alone. Together — on a job signed nine months ago with a 3% escalation assumption and a floating-rate facility — they are the difference between fee and loss. The work that pays this quarter is unglamorous: re-forecast committed jobs against actual July input indices, isolate diesel and electrical equipment as discrete risk lines, confirm long-lead release dates in writing before defending any 2027 milestone, and know which of your subcontractors carry data-center backlog. Cost forecasting, procurement visibility and disciplined change documentation are not overhead in this environment. They are the margin.
The data
Every figure traced to its primary release, with true observation period and release date. A complete metric-level ledger with URLs accompanies the edition.
Sources & methodology
- U.S. Census Bureau — Monthly Construction Spending, July 2026 (CB26-140), released Sept 1, 2026. source
- U.S. Bureau of Labor Statistics — The Employment Situation — August 2026 (USDL-26-1435), released Sept 4, 2026. source
- U.S. Bureau of Labor Statistics — Job Openings and Labor Turnover — July 2026, released Sept 1, 2026. source
- U.S. Bureau of Labor Statistics — Producer Price Index, July 2026 (as analyzed by AGC of America). source
- AGC of America — Construction input-cost and employment analysis of BLS data, Aug–Sept 2026. source
- Associated Builders and Contractors — Construction Backlog Indicator, July 2026 (released Aug 11, 2026). source
- Dodge Construction Network — Dodge Momentum Index, July 2026 (released Aug 6, 2026). source
- AIA / Deltek — Architecture Billings Index, July 2026. source
- Federal Reserve — FOMC statement and minutes, July 28–29, 2026; H.15 Selected Interest Rates. source
- U.S. Department of the Treasury — Daily Par Yield Curve Rates (2-yr, 10-yr), Sept 3–4, 2026. source
- U.S. Energy Information Administration — Gasoline and Diesel Fuel Update (retail diesel), Aug 21, 2026. source
- Federal Register — Further Adjusting the Tariff Regimes for Aluminum, Steel and Copper (Section 232), June 4, 2026. source
- Engineering News-Record — Employment subcomponent detail and equipment lead-time reporting, Aug–Sept 2026. source
Methodology. Monthly and quarterly series are identified by observation period, not the week of publication. Lead-time figures in Procurement are drawn from industry and supplier reporting rather than an official statistical series and are labeled accordingly. Metals PPI figures are the June reading as published; other PPI figures are July. Market-implied rate probabilities are derived from futures pricing, not from any Federal Reserve publication, and move daily. Where sources conflicted, the primary government release was used. Forward-looking statements are Syntecton judgment, labeled as interpretation, not data. This report is informational and is not investment advice.