2026 State of the US Construction Industry

Input costs moved one way and margins moved the other. US construction input prices are up 43% since January 2020, but average project margins have tightened from 5.8% to about 3.1% over the same period. In 2020, a GC could absorb a 6% cost surprise on a $50M job and break even. Today that same surprise wipes out the entire fee and puts the firm in the red. Volume does not cover mistakes when the margin is this thin.

Tariffs stopped being procurement noise and started deciding which jobs reach site. Quoted tariff rates on steel and lumber sit at 25% to 30% in 2026. The number that actually hits your job is different. Early 2026 AGC survey data put the realized cost impact at 14.1% on affected scopes, once you count re-quotes, supplier substitutions, and re-sequencing. That 14.1% figure is your real go/no-go threshold on tariff-exposed packages. If your contingency doesn't cover it, you're bidding to lose money. The commercial risk isn't just the material line. It's the knock-on delay and the extra general conditions you can't recover when your contract has weak change order provisions.

The dollars aren't spread evenly. Data centers and power infrastructure are absorbing a growing share of total construction put-in-place, while traditional non-residential sectors sit flat. Understanding where the spending actually lands changes how you staff, how you bid, and what equipment you commit.

The shift over six years is stark. In 2020, multifamily and commercial office dominated private-sector backlogs. By 2026, data centers and power generation have moved into the top spending slot, and IIJA-funded highway and water work fills the public side. If your firm is still chasing the same sectors it chased in 2021, you're fighting over a shrinking pool.

Fuel is back in the cost plan, even on projects that don't feel fuel-heavy. Crude petroleum is up about 20% year on year in 2026. That feeds diesel, asphalt, transport, and petrochemical products like PVC and polyethylene pipe. The cash risk sits between buyout and delivery. A spike in that window turns a "correct" estimate into a cash loss if you pay suppliers before you can certify the value on a pay app. The practical response: treat fuel-linked items as a managed exposure. Use indexed pricing clauses tied to a published index like EIA or PPI, and time-box subcontractor quotes to 30 days so you're forced to decide before the rate moves again.

GDP is tracking about 2.3% for 2026. That's enough activity to keep backlogs full but not enough to give owners pricing urgency. Owners are slower to issue notice to proceed, tighter on pay apps, and more willing to stretch payment terms past 60 days. The federal funds rate sitting at roughly 4.75%, versus 0.25% in 2020, makes every week of delayed payment expensive. At today's rates, carrying $2M in unbilled costs for 45 days costs you roughly $10,000 in financing alone. Contractors react by pushing risk down, often through longer retainage release and harder-to-agree change orders. The operator move is straightforward: score each job on cash strain before you price it. Look at front-end spend, long-lead procurement, and how quickly you can certify and collect.

The constraint stack for 2026 is clear: 25-30% tariffs, 43% cumulative input inflation, 4.75% cost of capital, and 2.3% GDP growth that won't bail you out with volume. Stop treating this like a pricing cycle. Treat it like an operating model change. Build tariff and fuel triggers into every buyout, time-box supplier quotes, and run weekly cost-to-complete on the packages that can hurt you fastest. Walk away from work where the payment terms and change order process make you the bank.

Tariffs don't just add cost. They break the link between your estimate and an orderable price. The hit lands in the gap between "budget pricing" and "supplier will hold it" pricing. That gap is where most contractors still carry assumptions, not commitments. Once trade policy moves, those assumptions become scope and schedule decisions made on the wrong numbers. The real question for 2026 procurement isn't whether tariffs raise prices. It's whether you lock, index, or defer each commodity, and what contract language survives owner and lender scrutiny when you do.

The first operational symptom is shorter quote validity and wider bid spreads on tariff-exposed packages. Subs stop holding quotes past seven to 14 days. That pushes risk premiums into every tier. Owners then see "contractor pricing up" and pause jobs. The job isn't always unaffordable because of the material. It's unaffordable because nobody trusts the number long enough to sign a fixed commitment. Each commodity has a different answer. Steel and copper, with 25-30% effective tariff rates and single-quarter PPI swings above 9%, need either a pre-buy commitment or an indexed clause. Lumber, at 31% above its 2020 baseline but less volatile quarter-to-quarter, can often tolerate a deferred buy with a price cap. Electrical switchgear and transformers, where lead times now stretch 40-60 weeks, require the earliest lock because the delivery slot matters more than the price.

A practical playbook sorts each material line into one of three buckets: lock, index, or defer. "Lock" means pre-buying with title transfer and insured storage once submittals are approved and the owner has released the spec. This makes sense for copper wire and bus bar, where tariff exposure is immediate and substitution options are limited. "Index" means tying the contract price to a named BLS PPI series with a defined measurement date, a cap, a floor, and a true-up frequency (monthly or quarterly). This is the right tool for structural steel and rebar, where price moves are large but trackable against a public index. "Defer" works for materials with stable domestic supply and lower tariff sensitivity, like softwood lumber, provided you set a ceiling price and a latest-commit date tied to your schedule. The key discipline: every line in your procurement plan needs an assigned bucket before the bid goes out, not after the owner signs.

Contract terms need to match the reality of volatility. A lump sum only works if you can either lock the supply chain or price a transparent adjustment. AGC's 2026 member survey found 38% of general contractors now include PPI-linked escalation clauses in bids as standard, up from 11% in 2020. The clauses that hold up under lender review share three features: a named BLS index series (not a vague "market conditions" trigger), a defined measurement window (typically comparing the PPI at bid date to PPI at purchase date), and a cap-and-floor band that limits owner exposure to, say, plus or minus 8-10%. ConsensusDocs 200.1 already includes a material price escalation rider. AIA A201 doesn't, so you need a supplementary conditions exhibit. Either way, the clause must name the index, the trigger threshold, the true-up timing, and who bears risk outside the band. Vague "equitable adjustment" language gets rejected by lenders and litigated by owners.

The labor shortage isn't a cost issue first. It's a delivery limit. Associated Builders and Contractors (ABC) puts the 2026 shortfall at 499,000 workers, up from roughly 320,000 in 2020. That gap turns schedule risk into commercial risk because you can't "buy" your way out at the last minute. If you don't have crews, you lose time. If you lose time, you lose overhead absorption and you invite liquidated damages.

Turnover is where the deficit becomes a half-million-dollar hole in your P&L. AGC's 2025 Workforce Survey reports roughly 73% skilled-trade turnover within 24 months. The replacement cost sits between $8,000 and $21,000 per worker once you count recruitment, onboarding, and lost output. Run that math across a 60- to 80-person field team and you hit the "half-million deficit" fast. Most firms bury it in general overhead, then wonder why jobs look fine on paper but cash is always tight.

The next hit is productivity. Paying more per hour is one problem. Getting less installed per hour is worse because it hides inside "normal" payroll. Average hourly wages rose from $28.50 (2020) to $39.80 (2026), a 40% increase. Over the same period, the productivity index dropped from 100 to 85. Combined, the effective cost per installed unit jumped roughly 64%. The estimator sees the first number. The project manager lives with the second. Neither realizes how fast the gap grows until the job is underwater.

Close the loop faster or you'll fund the overrun. The job doesn't fail when labor gets expensive. It fails when you find out too late that production has slipped. Firms that measure daily installed quantities against labor hours catch margin fade 18 days sooner than firms relying on monthly payroll reports. Four weekly leading indicators per project make the difference: crew stability by trade, overtime as a share of hours, foreman span of control (which grew from 1:8 to 1:12 between 2020 and 2026), and first-time quality by work package. Archdesk ties labor hours to quantities and cost-to-complete in one WIP view, so the first bad week shows up while you still have choices.

Construction is no longer one cycle. Capital is splitting between compute-and-power work and conventional buildings. Lawrence Berkeley National Laboratory data shows over 2,600 GW of generation and storage capacity sitting in US interconnection queues at end of 2024, nearly double the figure two years earlier. Owners are reserving grid positions years ahead, then pulling the construction program to match. If you bid this work, treat utility connection dates and long-lead electrical equipment as the real critical path, not structural steel or concrete pours.

Census Bureau "Value of Construction Put in Place" data confirms the split in dollars. Manufacturing construction spending ran above $230 billion annualized through late 2025, roughly triple its 2019 baseline. Nonresidential categories like office, retail, and lodging stayed flat or declined in real terms, growing around 2% year-on-year at best. Data center construction alone is projected to grow another 20% in 2026. Civil enabling works, power distribution, and industrial MEP scopes hold pricing power. Office fit-out and standard commercial interiors are fighting over fewer starts on harder terms. Stop using a single bid strategy across your whole pipeline.

Data center clients buy speed, but they also buy proof. Commissioning evidence moves from closeout to progress payment. QA paperwork becomes a commercial gate, not admin. An MEP subcontractor pricing a 50 MW data hall who buries testing and commissioning documentation costs inside "general conditions" will find those costs eat 2 to 3 points of margin when the client withholds payment pending documentation. Build the cost of evidence into your prelims, then tie each deliverable to a line on your schedule of values.

Backlog length tells you who can afford to say no. Dodge Data and Analytics reports backlog of roughly 10.9 months for GCs focused on data center work heading into 2026, versus 5.8 months for subcontractors in conventional fit-out and tenant improvement. A GC with 11 months of backlog can lock key subs early and walk away from one-sided payment terms. A fit-out sub with under 6 months takes fixed-price risk and 60-day pay cycles just to keep crews busy. AGC's 2026 outlook notes 61% of GCs report difficulty securing MEP subcontractors for conventional projects, up from 42% in 2022. The best mechanical and electrical trades are choosing longer, higher-certainty programs with steadier hours. Tag every job as "compute/power/infrastructure" or "conventional building," then set different rules for buyout timing, cash protection, and labor buffers. Archdesk shows margin, cash cycle, and resource load by sector tag in one view, so you place your best people and tightest commercial controls where the return is real.

Second-wave inflation shows up first in quote behavior, not headline indices. Suppliers are shortening validity and adding freight and fuel riders that sit outside the unit rate you thought you fixed. In early 2026, the BLS Producer Price Index (PPI) for ready-mix concrete was up about 6% year on year, and the PPI for truck transportation of freight was up about 7% over the same window. That hits every job, because concrete, aggregate, and deliveries sit inside most cost codes. Source: US Bureau of Labor Statistics, PPI series (2026).

The next hit is the gap between "award" and "orderable." A clean bid does not equal a basket of materials you can actually place purchase orders against. The Federal Reserve Beige Book through late 2025 recorded widespread reports of more frequent repricing and tighter terms for construction inputs. That turns buyout from a quarterly exercise into a weekly control process. If your estimating and procurement teams only refresh cost-to-complete at month end, the job carries the risk for weeks before anyone sees it in the numbers. Source: Federal Reserve Beige Book (late 2025).

Escalation clauses work when they are written as governance, not as a last-minute negotiation. A useful clause names the index, the basket, the baseline date, the measurement window, and the cap and floor. A weak clause says "market conditions" and guarantees an argument later. Industry standard forms now support the stronger approach. ConsensusDocs 200.1 includes an optional price adjustment exhibit that can be tied to published indices with a defined baseline. Source: ConsensusDocs 200.1 (price adjustment exhibit).

Procurement discipline is now a margin control system. A 2026 FMI procurement study found contractors using predictive allocation cut material cost variance to 1.8%. That doesn't come from better guessing. It comes from locking quantities and delivery windows early, then tying every commitment to a cost code and a contract mechanism. Source: FMI procurement study (2026).

Takeaway: build a one-page index schedule by package and agree it at award. Flow the same risk mechanism down into subcontracts and purchase orders, so you're not carrying risk you can't certify. Archdesk supports this by tying each buy to a cost code, a commitment, and a live forecast, so price movement shows up in the cost-to-complete the same week the PO changes.

Volatility doesn’t kill margin on its own. Slow decisions kill margin. Firms that hold profit in a choppy market run tight weekly control on three things: Work in Progress (WIP) integrity, forecast speed, and change control. They treat margin as something you manage during delivery, not something you measure at the end.

Cash makes volatility worse because it turns estimate error into interest cost. FMI benchmarking quoted in Draft 3 puts average days-sales-outstanding (DSO) for ENR Top 400 firms at 79 days in 2025, up from 67 days in 2020. That 12-day stretch means you carry more receivables for longer. Your overdraft or revolver does the heavy lifting, and finance costs creep into jobs that still look fine on a gross margin report.

Retainage is not just “money you get later”. It is a risk concentration point. On a live programme, retainage delays stack on top of long payment cycles, and both hit at the same time as closeout costs. CFMA’s 2025 survey cited in Draft 3 found 23% of firms had retainage disputes lasting longer than 180 days. That is long enough for a decent job to become a cash drain, especially if you have a few projects finishing together.

Change control is the margin lever most firms still under-run. CFMA benchmarking cited in Draft 3 shows top-quartile contractors recover 91% of qualifying change orders in the billing period they occur. Bottom-quartile firms recover 54%, often two to three cycles late. The difference is not contract knowledge. The difference is speed. If the event is not logged, coded, priced, and issued while the facts are fresh, you end up negotiating from memory after you have already paid the labour and the supply chain.

Run your margin protection like an operating rhythm, not a monthly finance task. Set a weekly rule for every live job: cost-to-complete updated, top change events priced and issued, and the biggest cash blockers named. Archdesk supports this by tying WIP, commitments, and change status to cost codes, so the problem shows up as a decision you can make this week, not a write-down you explain later.

Productivity did not “improve” between 2020 and 2026. The measurable win was stopping the slide while labour got tighter. The BLS multifactor productivity series for construction shows productivity falling by about 0.3% a year from 2020 to 2025. That is still negative, but it is a slower decline than the long-run average from 2007 to 2019, which sat closer to 0.5% a year. Treat that as the baseline reality. You are running the same jobs with less slack in labour, supervision, and programme float.

The most useful shift is how leading firms define “productivity”. They treat it as decision speed and work release, not how hard crews push. CII’s 2025 field benchmarking study tracked 430 US projects and found the biggest workflow gain came from compressing RFI turnaround. Adopting projects cut average RFI cycles from 11 days to 4.2 days. That difference is not admin polish. It is the difference between a crew staying on task and a crew standing down, then coming back out of sequence with a bigger knock-on effect.

Single “tools” did not move the needle on their own. The same CII study found a 12% composite productivity advantage only when firms combined three things, digital field reporting, automated look-ahead updates, and prefabricated MEP assemblies. Firms that adopted only one of the three showed no meaningful improvement. This matters commercially because half-adoption still costs time to run. It adds meetings, double-entry, and handoffs. If you cannot connect field progress, constraints, and procurement promises into one weekly rhythm, the overhead cancels out the gain.

Prefabrication is where scarcity turns into a real output advantage, but only when you plan it early enough to protect the supply chain. The Mechanical Contractors Association of America’s 2025 membership survey found prefabrication made up 22% of installed MEP scope on projects over $50m, up from 11% in 2020. Projects at 30% or more prefab share reported 18% fewer field labour hours per installed unit. The trap is treating prefab as a site rescue plan. The firms getting the hours back planned prefab packages eight to ten weeks ahead of install, and locked materials early so the shop did not become another constraint.

Run one change next week. Put RFI and submittal ageing on the same weekly ops agenda as labour and programme, with named owners and due dates. Then add a constraint log tied to next week’s planned work, with a simple Friday measure, what got cleared and what did not. Archdesk supports this operating rhythm by keeping RFIs, procurement promise dates, and cost codes in one place, so you can see which package is about to idle a crew before the labour hours are already spent.

“Fixed price” still dominates on paper, but the real shift is inside the clauses. A 2025 CFMA benchmarking study found 62% of awards by value are still classed as fixed-price. Most of those deals now carry escalation riders, allowance schedules, or contingency pools. Treat them as hybrids. Your margin depends on whether your team can run the admin that makes the hybrid terms pay.

Contingency is no longer a “site buffer”. It is a governed fund with sign-off levels. A 2025 Deloitte survey found 71% of institutional owners now require a formal drawdown protocol with thresholds and approvals. That changes how you run jobs. If you do not tag each drawdown to a cost code and a clear event, you will not get agreement when it matters. You will get a dispute at closeout, which is the worst time to be proving facts.

Escalation clauses fail for boring reasons. The clause is only worth what you can evidence in days, not weeks. ConsensusDocs 200.1 includes an optional price adjustment exhibit that ties changes to published indices with a defined baseline. That is the level of discipline owners now expect, even under other forms of contract. Your team needs a simple pack for any claim: baseline quote, quote expiry, buy date, index reference, and the calculation. If one item is missing, the commercial argument collapses.

Self-perform and direct supply agreements are being used as contract controls, not growth plays. ENR’s 2026 survey reports 34% of firms in the $100m to $500m revenue band now self-perform at least two trades, up from 19% in 2021. Those firms also reported tighter cost variance on self-performed scopes, 2.1% on average versus 5.7% for fully subcontracted packages. The commercial win is not just markup saved. It is fewer expiring quotes and fewer gaps between what was priced and what can be ordered.

Controls break when your records live in different places. Escalation evidence fails if the baseline quote is in an email thread, the buy date sits in a spreadsheet, and the cost code is in the accounting system. Contingency governance fails when nobody can match a drawdown request to live commitments and a current forecast. Archdesk closes that gap by tying quotes, purchase orders, commitments, and WIP to the same cost codes, so commercial rules become enforceable on the day the cost hits.

Practical move for next week: build a “contract proof pack” for your top 15 cost lines. For each line, store the baseline quote and expiry, the buy date, the named index and rule, and the notice window. Then run one live test on a current job. Ask your team to produce the pack inside 48 hours. Fix whatever stops them. That is the difference between having protection in the contract and getting paid under it.