Quanta just booked the AI buildout's labor constraint as a $63.7 billion backlog
Quanta Services, the contractor whose crews build the power lines data centers run on, posted revenue up 57 percent, a record $63.7 billion backlog, and raised guidance on every metric. By operators' own account the buildout's binding constraint is skilled labor, and that is the part Quanta sells.
Richard Tang · 3 min read
Quanta Services+6.02% — Quanta Services, up 6.02 percent today just turned the AI buildout's constraint into a line item. The Houston contractor, whose self-perform crews build the transmission lines and substations that data centers run on, reported third-quarter revenue of $12.01 billion, 57 percent above a year earlier 1. Adjusted earnings of $6.06 a share beat the $5.00 consensus, the sixth beat in a row 2.
The most telling number is the order book: a record $63.7 billion of backlog, $41.8 billion of it under firm contract 1. PWR is the listed proxy for the buildout's labor constraint, and it sits at the front of the payment queue, ahead of the chip.
The backlog is the capex wave with signatures on it
The money trail runs from ratepayers and hyperscalers, through utility capital plans, into Quanta's contracts. A year ago the total backlog stood at $39.2 billion 1. Not all of it is AI by any means: management put data centers at roughly a tenth of the business on its year-end call, with the rest spread across grid reliability, power generation and industrial customers 4.
The AI wave reaches Quanta mostly as load growth that forces utilities to rebuild the grid. That is the durable version of the exposure, because the spend is contracted and regulated rather than tied to any single hyperscaler's quarter.
Total backlog climbed from $39 billion to $64 billion in twelve months
Data
| Total backlog | |
|---|---|
| Q3 2025 | $39.17B |
| Q4 2025 | $44B |
| Q2 2026 | $53.4B |
| Q3 2026 | $63.73B |
The bought part of the growth is the smaller part
The skeptic's case is that the growth was acquired, and it is partly right: five acquisitions closed in the first nine months of 2026 1. The bought share is measurable, though. Zacks put this year's acquired businesses at $1.2 to 1.4 billion of 2026 revenue 3.
The full-year revenue guide now sits at $41.3 to 41.6 billion, raised twice since a September outlook of $39.3 to 39.7 billion 13. The quarter's own split between organic and acquired is not disclosed 1, but the step in the revenue line began before most of those deals closed 9.
Quarterly revenue has doubled in three years, with the steepest step this year
Data
| Quarterly revenue | |
|---|---|
| Q3 '23 | $5.62B |
| Q4 '23 | $5.78B |
| Q1 '24 | $5.03B |
| Q2 '24 | $5.59B |
| Q3 '24 | $6.49B |
| Q4 '24 | $6.55B |
| Q1 '25 | $6.23B |
| Q2 '25 | $6.77B |
| Q3 '25 | $7.63B |
| Q4 '25 | $7.84B |
| Q1 '26 | $7.88B |
| Q2 '26 | $9.56B |
| Q3 '26 | $12.02B |
Buying the bottleneck, and carrying the fixed costs
Quanta is also buying the bottleneck itself. It just paid about $245 million for an energy-services firm and a steel monopole maker, pulling fabrication in-house 1, on top of an earlier $500 to 700 million investment in transformer manufacturing 4. It issued $2.00 billion of senior notes in August and still prepaid its term loan 1.
The risk is that crews and factories are fixed costs, and a backlog converts to revenue only while the capex keeps running. The same print answers it: guidance raised across every metric, and chief executive Duke Austin pointing to strong double-digit earnings growth again in 2027 1.
The limiting link is trained crews, and crews can be trained
Demand is no longer the constraint; the industry says so itself. Ninety percent of data center operators call staffing shortages a critical restraint on building, and Kelly Services' verdict is blunt: the biggest constraint "is no longer just power, land or equipment. It's talent" 5.
Data center job postings have more than doubled in two years while overall postings fell about 12 percent 7. Quanta's moat sits exactly there, in what it calls the largest craft labor force in North America, built across more than 300 operating companies with its own training network 6.
The open door into the AI economy is a skilled trade
The gap between a model that exists and a data center that runs is closed by people who can hang wire, and Quanta trains its own. That is where this buildout's reward is going, and it is where it belongs. For anyone asking what to learn next, the postings data gives a plain answer: the open door into the AI economy is a skilled trade.
Get Power by email, free
How AI's hunger for electricity moves utilities, power plants and the grid. In our daily email, every weekday.
