This group sells the act of building data centers and the power/industrial plants that feed them. The "product" is not a thing you can hold — it is engineering, procurement, and construction labor (this bundle is called EPC: engineering, procurement & construction — designing a project, buying its parts, and physically building it). The companies here — Quanta Services (PWR), EMCOR (EME), Comfort Systems USA (FIX), Sterling Infrastructure (STRL), and AECOM (ACM) — are the general contractors, electrical/mechanical specialists, and program managers who turn a hyperscaler's spending plan into a finished, powered building full of running servers. The buyers are the people building the AI cloud: hyperscalers (the largest cloud operators — Amazon, Microsoft, Google, Meta), colocation landlords ("colos" — companies that build data centers and rent space/power to others), and the utilities and pipeline owners who must add electricity to serve them.
Demand for this work is running ahead of the supply of people who can do it: hyperscaler and colo spending is climbing while the trades needed to build are in chronic shortage, so backlogs (signed future work not yet billed) sit at record highs, per the scan. The limit on supply is not steel or capital — it is skilled hands and qualified project managers, which take years to train. What the market charges today for a dollar of this group's earnings is laid out in the Price section as a range, with no judgment attached; the reader does the weighing.
Source: 500-stocks scan, "Engineering & Construction Firms" section (/Users/ravf/projects/work/.claude/worktrees/sector-hub/research/investments/500-stocks/06-industrial-construction.html); demand-pool dollar figure is in the scan, the growth rate is a general-knowledge estimate (see note in Demand).
The product is built capacity — a finished data-center campus, substation, gas plant, or transmission line, delivered ready to run. It is sold per project, priced in dollars of contract value, and a single large AI campus is a multi-year, multi-billion-dollar job. Within that, each company sells a slice of the build:
How the cash actually arrives: a customer signs a contract, the work goes into backlog (signed but not-yet-performed work — the closest thing this industry has to a forward order book), and the company bills as it builds, collecting cash over the life of the job. The owner economics are capital-light relative to the customer: these firms do not own the data centers; they sell labor and project management and earn a margin (operating margin = operating profit as a percent of revenue; for this group typically single-digit to low-double-digit) on a large revenue base. The scarce resource they monetize is a trained, deployable workforce, not factories.
Source: 500-stocks scan, E&C section; product/segment framing grounded in standard company business descriptions (10-K segment definitions).
Today (closest to contracted): the scan states hyperscalers and colocation providers are spending $100B+ per year on US data-center construction, and that backlogs at major E&C firms are at all-time highs backlog = signed work. Backlog is the most concrete demand signal in this group because it is money customers have already committed; the scan's "all-time high" language indicates demand that is contracted, not merely hoped for.
Forward (forecast): reasoning from the premise that AGI is arriving, demand points one direction. As AI capability compounds and compute/physical-AI demand rises, the buyers are racing to add compute faster than power and buildings can be delivered. Every incremental gigawatt of AI compute requires not just the data center shell but new substations, transmission, and often dedicated generation — all of which is this group's work. So demand here is leveraged to two stacked build-outs at once: the data centers themselves and the electrical grid expansion to power them. forecast
Who the buyers are, and why they may be durable: the spending is concentrated among a handful of cash-rich hyperscalers (Amazon, Microsoft, Google, Meta) plus colos (e.g., Equinix, Digital Realty, and private operators) and the regulated utilities expanding to serve them. These buyers fund construction from operating cash flow and large balance sheets, which can make this demand less fragile than typical construction cycles tied to interest rates or consumer spending. est. (buyer-list and balance-sheet characterization are general knowledge, not a live figure)
✓ VERIFIED — the following figures were confirmed from primary sources after initial publication:
Source: 500-stocks scan E&C section ($100B+, all-time-high backlogs); growth rate and forward sizing are general-knowledge estimates; AGI demand framing is forecast.
"Supply" here is how much building work the industry can actually deliver per year, and the scan is blunt about the constraint: a severe skilled-labor shortage. Electricians, ironworkers/pipefitters, and experienced project managers are in critically short supply; the scan states firms cannot hire fast enough to match backlog growth and calls this a multi-year constraint with no quick fix.
Why the scan frames labor — not steel or money — as the binding constraint:
Market-share structure (who controls supply): this is a fragmented industry with a handful of scaled players on top. No single firm dominates US data-center construction; the named public companies are among the larger players in their niches — Quanta (PWR) in electric power/transmission, EMCOR (EME) and Comfort Systems (FIX) in mechanical/electrical building services, Sterling (STRL) in site/civil e-infrastructure, AECOM (ACM) in design/management. Their stated edge is scale of trained workforce + safety/execution record + ability to self-perform (do the work with their own crews rather than subcontract), which can let them win the largest, most complex jobs that smaller contractors cannot staff. est. (relative positioning is general knowledge, not a live share table)
Source: 500-stocks scan E&C "Supply Constraints" (skilled-labor shortage, multi-year, no quick fix) and adjacent material sections (steel/concrete rated "medium"); share/positioning is general-knowledge estimate. Note: STRL is not named in the scan's E&C company list (see Players/Sources).
Put the two sides together and the product is, on the scan's facts, structurally short: contracted demand (record backlogs, $100B+/yr spend) is rising faster than the industry's ability to deliver it (labor-capped). The evidence is the backlog itself — work is being signed faster than it can be built, which is why backlogs are at all-time highs rather than being worked down. When a contractor's order book grows while it is straining to hire, that is the arithmetic signature of a supply-short market.
| Factor | Demand side (the work) | Supply side (ability to deliver) |
| Direction | Rising — AI data centers + grid build | Rising slowly — labor-capped |
| Hard signal | Backlogs at all-time highs contracted | Cannot hire fast enough (scan) |
| Binding limit | Hyperscaler/colo budgets (large, funded) | Skilled trades + project managers |
| Time to relieve | Years (AGI build-out, forecast) forecast | Years (apprenticeships, experience) |
| Net | Short — demand > deliverable supply | |
When could it flip to oversupply? Two paths, both forecast forecast: (1) demand falls — hyperscalers collectively pause AI capex (capex = capital expenditure, the money spent building long-lived assets; a collective pause is sometimes called an "AI capex hangover"), which would shrink new awards and let the labor pool catch up; or (2) supply catches up — sustained training/immigration of trades plus productivity tools (including AI-assisted project management and, eventually, more automated construction) raise deliverable capacity. Under the premise that AGI is arriving, the demand path is more likely to keep widening the gap than to close it in the near term, because each new tier of compute pulls forward both data-center and grid construction. The factual caveat: this has historically been a cyclical industry, and a capex pause is the single biggest risk to the "short" condition. The gap is observable today; its durability is a forecast, not a fact.
Source: 500-stocks scan (backlogs all-time high; labor shortage multi-year); flip scenarios and durability are forecast / general reasoning under the AGI premise.
The table below bridges to individual deep-dives. Revenue and market-cap figures are rough, rounded, and drawn from general knowledge of recent filing levels; treat them as directional, not live-verified est.. "Exposure to THIS product" means how much of the company's revenue is tied to data-center / power-infrastructure construction versus other end markets.
| Company (ticker) | What it builds | Exposure to data-center/power build | Rough size | Position / edge |
| Quanta Services (PWR) | Electric power, transmission, substations, on-site power; some communications & pipeline | High and rising — power-infrastructure is the core; benefits from both data-center and grid build est. | ~$25B+ revenue; large-cap est. | Among the largest US electric-power specialty contractors; large self-perform labor force — the scarce skill est. |
| EMCOR Group (EME) | Mechanical & electrical building services, construction, facilities | Meaningful — data centers a growing slice of a diversified non-residential mix est. | ~$14-15B revenue; large-cap est. | Scale + diversification across many building types est. |
| Comfort Systems USA (FIX) | HVAC, mechanical, electrical install & service for buildings | High and rising — data-center/advanced-tech is a growing share of mechanical work est. | ~$7-8B revenue; large/mid-cap est. | Among the most exposed in the group to data-center cooling/mechanical density; record backlog est. |
| Sterling Infrastructure (STRL) | Site development / civil ("e-infrastructure"), foundations, low-voltage | High in its e-infrastructure segment — large data-center site-work content est. | ~$2-2.5B revenue; mid-cap est. | Smaller-cap name with concentrated data-center site-prep exposure; higher-margin e-infra mix est. |
| AECOM (ACM) | Design, engineering, program/construction management | Lower per-dollar — design/management is a thin slice of total project cost; diversified across infrastructure est. | ~$16B revenue (much of it pass-through); large-cap est. | Professional-services model; lower margin, less labor-trade exposure, more diversified est. |
The scan's broader E&C list names EME, FIX, PWR, ACM plus more diversified or different-end-market firms — MasTec (MTZ), Dycom (DY), Jacobs (J), Fluor (FLR), KBR (KBR) — which are comparators less concentrated on this specific data-center build. Sterling (STRL) is not in the scan's E&C list; it is included here from general knowledge as the clearest small-cap data-center site-work name, and its inclusion is therefore an estimate, not grounded in the scan est..
Source: 500-stocks scan E&C company list (EME, FIX, MTZ, PWR, DY, J, FLR, KBR, ACM); STRL added from general knowledge (not in the scan); revenue/cap magnitudes are general-knowledge estimates from recent filing levels (not live-verified).
In plain money terms, here is roughly what the market charges to own a dollar of this group's economics. One common yardstick is price per dollar of this year's earnings (the P/E multiple — market value divided by annual net profit; "you pay $X of market value for $1 of yearly profit") and another is price per dollar of revenue (price-to-sales; relevant because these are thin-margin businesses where revenue scale matters). All figures rough and not live-verified est..
Money-in / money-out shape: this group is capital-light for the contractor — the customer funds the buildings, the contractor funds mostly working capital and people. That means owner cash (free cash flow — operating cash left after the spending needed to keep the business running) can be a meaningful fraction of profit, but it swings with the timing of billings on big jobs (a contractor can be cash-hungry while a backlog ramps, then cash-generative as it bills out). The cash engine is margin on a large, fast-growing revenue base, not return on owned assets. So what an owner is paying for reduces to one question: will the record backlog and the labor-scarcity condition convert into rising profit per share for years? If the gap (demand > deliverable supply) persists, today's multiple is paying for growth that is partly contracted (backlog) and partly forecast (AGI build-out). If a capex pause hits, the same multiple is paying for growth that may not arrive. Those are the two sides of the arithmetic; the reader weighs them.
Source: valuation multiples and FCF characterization are general-knowledge estimates (not live-verified); capital-light/working-capital dynamics reflect the industry's standard contracting model.
Where a company-level deep-dive would add the most factual information, organized by role in this group (neutral pointers, not recommendations):
An order that surfaces the most information per hour: FIX or STRL first (most concentrated signal), then PWR (labor-slice exposure + grid leverage), then EME/ACM (diversified comparators). This is a learning order, not a ranking of attractiveness.
Source: role assignment derived from the scan's segment descriptions and general business knowledge; not a buy/sell ranking.
What was used:
Hard vs approximate (plainly):
Source: /Users/ravf/projects/work/.claude/worktrees/sector-hub/research/investments/500-stocks/06-industrial-construction.html (scan); general knowledge (cutoff ~early 2026) for non-grounded figures, all labeled approximate and not-live-verified.