How each claim is supported:
Tap any marked phrase for its source.
Section 1 of 9
How this industry makes money
In one sentence
A contractor wins work by bidding, then earns the difference between the bid and what the job really costs. A homebuilder buys land, builds homes and sells them, keeping what is left after land, construction and sales costs.
Six pools of money decide the year.
| Pool | What it is | Who controls it |
|---|---|---|
| Backlog | Work won and not yet built | Owners and buyers, and how well the firm bids |
| Job margin | The gap between the bid and the real cost | Estimators, project managers and change orders |
| Labor | Craft workers and supervisors | The firm, within a tight labor market |
| Materials | Steel, lumber, concrete, equipment | Suppliers and distributors |
| Overhead | Office staff, estimating, insurance | The firm |
| Cash timing | When owners pay, and retainage held back | Owners and contracts |
For homebuilders, land and mortgage rates matter most. Lennar's margin fell as prices dropped and land cost more, even though construction costs fell.
AI touches three of these most: job margin, through estimating; overhead, through paperwork; and backlog, through faster bids. Labor is the biggest constraint, and AI touches it least.
Section 2 of 9
Where AI touches the economics
In one sentence
Demand: the data center boom. Data centers and power projects are where contractors expect growth. Record data center construction is under way. Firms that can do this work are busy; others face slower markets.
Estimating: faster bids. AI can read drawings, count materials and draft estimates. Estimating is one of the most common uses of AI among contractors. A faster estimate lets the same team bid more jobs. A wrong estimate wins jobs that lose money.
Overhead: less paperwork. Office and administrative work is the most common use of AI in construction. Contracts, submittals, meeting notes and invoices take less time.
Planning: fewer surprises. AI can check schedules and flag conflicts before work starts, if the project data is current. Fewer than half of firms in one survey said site teams have real-time data.
Labor: the limit. More than four in five contractors struggle to hire craft workers. AI does little for the people who do the building.
Time saved is not money saved. An estimator who bids faster earns the firm more only if the extra bids win good jobs at good margins.
Section 3 of 9
What is already happening
In one sentence
| Who | What they did | Reported result | What the source does not show |
|---|---|---|---|
| US contractors (AGC and Sage, 2026) | Adopted AI for office work, estimating and preconstruction | 61% use AI or plan to spend more, up from 44% | What it saved or earned |
| US contractors (AGC and Sage, 2026) | Rated the year's markets | Data centers net +57, the only double-digit gain; power +34 | Which firms can win that work |
| North American data center builders (CBRE, 2026) | Built for AI demand | A record 7,481 megawatts under construction, more than 80% preleased | How long the boom lasts |
| Asia Pacific construction firms (Deloitte and Autodesk, 2026) | Surveyed on digital tools | 46% use AI, up from about a quarter in 2023; a quarter still run mainly on paper | US figures on the same basis |
| Lennar (2026) | Sold homes in a slow market | Gross margin fell from 18.7% to 15.2% | Any effect of AI, which the source does not mention |
| Construction leaders (Deloitte summary, 2026) | Surveyed on AI | About three in four expect AI to help; about two in three do not use it in operations | Which firms are ahead |
Two patterns stand out. AI's biggest effect on construction so far is as a customer: data centers. And as a tool it is still mostly in the office, while the jobsite runs much as before.
Section 4 of 9
What could change next
In one sentence
Scenario 1: The data center boom slows. AI companies cut spending, and data center projects are delayed or cancelled. Assumes: AI spending growth slows. Signal to watch: data center projects paused, and preleasing of new capacity falling below its current level. Last checked: September 2026. More than 80% of capacity under construction is preleased, and contractors expect data centers to grow more than any other market.
Scenario 2: AI estimating becomes standard. Most contractors use AI to count materials and draft bids. Assumes: the tools prove accurate on real projects. Signal to watch: owners asking bidders how they used AI, or bid turnaround times shrinking across the market. Last checked: September 2026. Estimating is already one of the top AI uses among contractors.
Scenario 3: AI reaches the jobsite. AI helps with schedules, safety and progress tracking on site, not just in the office. Assumes: site teams get reliable, real-time project data. Signal to watch: the share of firms whose site teams have real-time data rising well above half. Last checked: September 2026. Fewer than half of firms say site teams mostly have real-time data.
All three reward the same preparation: a clear view of how much of your backlog depends on data centers, estimating tools tested on past bids, and project data your site teams can reach.
Section 5 of 9
Who gains and who loses
In one sentence
Gain: contractors who can build data centers and power projects. Demand there is the strongest in the industry.
Gain: firms whose estimators use AI and check it. They bid more jobs with the same team.
Gain: specialty trades in short supply, such as electrical and mechanical, whose skills data centers need.
Lose: firms that bid faster without checking. They win the jobs they priced too low.
Lose: contractors in cooling markets. Several markets expect less work this year.
Mixed: homebuilders. AI can help with plans, paperwork and sales, but their margins depend on land and rates.
Section 6 of 9
Where to look before you invest
In one sentence
| Where | Prerequisite | Cost range for a firm with $20 million to $1 billion in revenue | Most common reason it fails | Pitch you will hear |
|---|---|---|---|---|
| Office and paperwork | A list of the documents that take the most time | Low; often part of existing tools | Nobody checks what the AI drafts | "Never write a submittal again" |
| Estimating help | Past bids with real quantities and costs to test against | $20,000 to $150,000 a year | Missed quantities win jobs priced too low | "Bid twice as many jobs" |
| Project data your site teams can reach | Plans, schedules and changes in one place | $50,000 to $300,000, plus training | Site teams keep using paper | "One source of truth" |
| Schedule and safety checks | Current project data and a person who acts on the flags | Varies widely; test on one project first | Alerts nobody reads | "Predict delays before they happen" |
The cost ranges are illustrative and will be replaced with reviewed figures. The order matters more than the amounts: office work first, estimating tested on past bids second, jobsite tools only once the data is ready.
One test before any contract. Ask the vendor to run its tool on three past bids where you know the real quantities and final costs.
Section 7 of 9
Up and down the chain
In one sentence
Owners and developers. Demand for data centers comes from AI companies and the landlords who lease to them.
Suppliers. Distributors of building materials are using AI to price and handle orders.
Equipment and materials makers. Makers of steel, electrical gear and cooling equipment are ramping up for data centers.
Lenders and investors. Funds and lenders are deciding which projects to back based on what AI will do to demand.
Section 8 of 9
Questions for your team
The same ground, for whoever owns each answer. Bring them to your next leadership meeting.
Ask leadership
How much of the backlog depends on data center and power projects, and what if they slow?
A good answer
A share of backlog and a plan if it falls.
Push back if you hear
“Data centers are booming.”
Ask chief estimator
Has the AI estimating tool been tested on past bids with known costs?
A good answer
Results from past bids, with errors by trade.
Push back if you hear
“It is faster.”
Ask project executive
Can site teams see current plans, schedules and changes?
A good answer
Yes, and a measure of how often they use it.
Push back if you hear
“They call the office.”
Ask finance
Are margins on jobs won with AI-assisted bids holding?
A good answer
Margin by job, AI-assisted against not.
Push back if you hear
“It is too early to tell.”
Ask head of people
Do you have the craft workers to take on data center work?
A good answer
A headcount plan by trade.
Push back if you hear
“The subs will find them.”
Ask risk manager
Who checks every AI-drafted contract, submittal or safety plan?
A good answer
A named reviewer for each.
Push back if you hear
“The tool is accurate.”
Section 9 of 9
What changed
In one sentence
September 29, 2026: first published.
August 2026 (reflected at launch): a record amount of data center capacity was under construction, most of it preleased. Affects sections 2, 3 and 4.
2026 (reflected at launch): Lennar reported its gross margin falling to 15.2%. Affects sections 1 and 5, and supports the guide's point that homebuilders' pressures are about land and rates, not AI.
January 2026 (reflected at launch): 61% of contractors use AI or plan to spend more, and data centers lead all markets. Affects sections 1 through 5.
Decision exercise
Write your prediction and one line of reasoning before you compare it with the guide.
Sources
Associated General Contractors of America, Contractors have dampened expectations for 2026, apart from data centers and power projectsJanuary 8, 2026
Associated General Contractors of America, Data DIGest, January 5-9, 2026January 2026
CBRE, North American data center demand continues to outpace supply despite record construction activityAugust 27, 2026
Deloitte and Autodesk, State of Digital Adoption in the Construction Industry 20262026