The Four Layers of the Build
These four names get lumped together as "AI-infrastructure contractors," but they don't do the same job. They sit at different layers of the same construction stack, and the differences are exactly what separate the winners from the also-rans when the cycle gets tested.
- STRL — Sterling Infrastructure. The dirt. Site development and civil work: grading, foundations, the massive concrete pads a data center sits on. Sterling's E-Infrastructure segment does the earthwork before anyone installs a single server. First trade on site, first to feel a slowdown.
- FIX — Comfort Systems USA. The systems (specialist). Mechanical, electrical and plumbing (MEP) — the cooling, power distribution and piping inside the building. Increasingly a modular-prefab shop, building electrical and mechanical assemblies in a factory and shipping them to site. Data-center/tech is now ~45% of revenue.
- EME — EMCOR Group. The systems (scale + diversified). The largest US mechanical/electrical contractor. Same MEP work as FIX but bigger, plus a large building-services arm and exposure to healthcare, manufacturing, government and network/communications. The diversified version of the same trade.
- PWR — Quanta Services. The power. The grid itself — high-voltage transmission, substations, "speed to power" interconnection. You can pour every pad and hang every chiller you want; without Quanta's layer the building never energizes. This is the genuine bottleneck of the entire buildout.
Read that list again and the investment question reframes itself: you're not choosing the "best contractor." You're choosing which layer of the build has the most durable demand and the most defensible economics — and paying the right price for it.
Where the Prices Are
All four are in the same technical condition: below the 50-day, RSI in the high-30s to low-40s (near oversold, not yet turning), sitting on or just above the 200-day. This is a synchronized momentum unwind, not a single-company problem. As of the August 24 close:
| Ticker | Price | Off 52wk high | vs 50-day | vs 200-day | RSI(14) |
|---|---|---|---|---|---|
| STRL | $496.66 | −50.6% | −26% | −4% | 36.3 |
| FIX | $1,609.69 | −22.4% | −9.5% | +10% | 41.1 |
| EME | $764.90 | −19.6% | −4% | +1.4% | 42.2 |
| PWR | $616.78 | −21.8% | −8% | +6.3% | 39.7 |
Notice STRL is the outlier — a 50% haircut versus roughly 20% for the other three. That's the market treating the dirt layer as the most cyclical and the most richly priced going in. Hold that thought; it's central to the verdict.
What Q2 Actually Said
The fundamentals are not ambiguous. Every one of these was a record quarter with raised guidance. Side by side, Q2 2026:
| STRL | FIX | EME | PWR | |
|---|---|---|---|---|
| Revenue | $1.17B | $3.3B | $5.15B | $9.56B |
| Revenue growth (YoY) | +90% | +50% | +20% | +41% |
| EPS growth (YoY) | +116%* | +92% | +35% | +71%* |
| Backlog | $5.6B† | $14.1B | $17.1B | $53.4B |
| Margin profile | ~22% EBITDA | ~15% EBITDA | 10.6% op | ~11% EBITDA |
| Fwd P/E (approx) | ~25x* | ~44x | ~23x | ~37x |
*STRL and PWR EPS/valuation are on an adjusted basis; STRL's GAAP P/E screens far higher (~60x) because of acquisition amortization — the optical trap discussed below. †STRL combined backlog including unsigned awards; signed backlog is $4.33B.
A few things jump off that table. PWR's $53.4B backlog dwarfs the group — that's what the grid bottleneck looks like on a balance sheet. STRL has the fastest growth and the fattest margins (site work at 22% EBITDA is genuinely impressive for civil construction), but the smallest backlog and the most acquisition-dependent earnings. EME grows the slowest on revenue but converts it into a 35% EPS gain and the cleanest valuation. FIX is the momentum darling — and the most expensive thing in the room.
Why the Whole Group Is Selling Off
If the quarters were this good, why is everything red? Three forces, in order of importance.
1. AI-capex digestion fear. This is the one that matters. Every dollar of these backlogs traces back to hyperscaler capital spending — Microsoft, Amazon, Google, Meta, OpenAI and the neoclouds. The market is now pricing a non-zero probability that 2026 is the peak rate of change in that spending. Backlog is a lagging comfort but a leading tell: the day bookings stop growing, these multiples compress violently. The stocks are front-running that risk regardless of what this quarter printed.
2. A momentum unwind. These names went vertical — FIX is up over 900% in three years, STRL nearly quadrupled before this pullback. When money rotates out of momentum, the highest-multiple names give back the most. That's mechanical, not fundamental, and it's why the expensive names can keep falling even on good news.
3. The skilled-labor ceiling. This is the real-world constraint, and anyone who runs field crews knows it: a record backlog is worthless if you can't staff it. Every one of these companies is fighting for the same electricians, pipefitters and operators. Wage inflation is the margin risk that doesn't show up until the job is underway.
Head to Head: FIX vs EME
The two MEP names are the cleanest direct comparison — same core trade, opposite profiles. Worth settling before the verdict.
FIX is the better business; EME is the better stock right now. Comfort Systems is arguably the highest-quality operator in the entire group — gross margins cleared 25% for the first time in company history, free cash flow is prodigious, and the modular-prefab strategy is a genuine structural edge that turns messy on-site labor into repeatable factory output. But it trades at ~44x forward. In a tape that is actively de-rating growth, the most expensive name has the most multiple to lose, and its beta cuts both ways.
EMCOR does the same work at nearly half the multiple (~23x), grows EPS at a very respectable 35%, has beaten estimates eight quarters running, and — critically — is the diversified one. Data centers are a big piece, but healthcare, manufacturing, government and network services are large enough to cushion a hyperscaler air pocket. In an environment where the single biggest fear is exactly that air pocket, EME's boring diversification is a feature, not a bug.
The Verdict — Who I'd Back Right Now
Being forward-looking about this means asking not "who had the best quarter" (they all did) but "who has the most durable demand at the least demanding price, given that the thing everyone's afraid of is a capex slowdown." On that test, the ranking is clear.
My pick for current conditions: EMCOR (EME). It's the best risk-adjusted way to own this theme today. Cheapest of the quality names, most diversified (the direct hedge against the one headwind pressuring the whole group), the most consistent execution record, and a record $17B backlog that isn't riding on a single customer set. It won't be the biggest winner if the buildout re-accelerates — but it's the one that holds up best if the market's capex fear turns out to be right, and it still participates fully if it's wrong. That asymmetry is what you want after a drawdown you can't precisely time.
Very close second, and the better pure-thesis bet: Quanta (PWR). If you believe the AI buildout has years left — and the grid interconnection queues say it does — Quanta owns the actual bottleneck. Grid and transmission demand is utility-driven, not purely hyperscaler-driven, which makes it structurally less cyclical than the building trades: the grid needs upgrading whether or not the next data center breaks ground on schedule. A $53.4B record backlog, a Moody's upgrade, and 41% organic growth back it up. At ~37x it's not cheap, but it's the highest-conviction "the buildout is durable" expression on the board. For a longer holding period, PWR may well beat EME.
The high-torque rebound trade: Sterling (STRL). This is the one in the wheelhouse of anyone who likes asymmetry. A company growing revenue +90% with 22% EBITDA margins, cut in half, now trading around 25x forward adjusted earnings. If the buildout continues, that's the most violent snap-back candidate here. But it carries the most risk: the most hyperscaler-concentrated demand, the smallest backlog, earnings leaning on acquisitions, a leadership transition, and — as the dirt layer — the first to feel any slowdown. High reward, highest risk. Size it like the trade it is, not the core position.
Best business, wrong price for this tape: Comfort Systems (FIX). Nothing here is a knock on the company — it's elite. But ~44x into a de-rating momentum unwind is the worst entry-risk of the four. I'd want it lower, or I'd want confirmation the group has stopped falling, before paying up.
The Headwinds That Could Break This
Every one of these theses dies the same way, so be honest about it:
- A hyperscaler capex pause. The whole group is a leveraged bet on four or five companies' spending plans. If any of them signals digestion, backlogs stop growing and the multiples reset again — the diversified (EME) and utility-driven (PWR) names bruise; the concentrated (STRL) and expensive (FIX) names break.
- Labor and margins. The skilled-trades shortage is the binding constraint on converting backlog to earnings. Wage inflation or a botched fixed-price job can turn a record backlog into a margin miss.
- Integration risk. STRL and PWR are both growing partly through acquisition. Contractors lose money on integration far more often than on demand.
- Rates and permitting. Higher-for-longer pressures the multiples directly, and grid interconnection/permitting delays can push project timelines right — a double-edged sword that boosts PWR's demand while also gating the whole theme's pace.
- The valuation itself. Even after these haircuts, only EME looks genuinely reasonable. If the market decides cyclical contractors shouldn't carry 35–45x multiples at all, there's more air underneath the group.
How that changes the decision: it's precisely why the pick is EME over FIX and the pure-thesis nod goes to PWR over STRL. When the dominant risk is a demand-concentration shock, you lean toward diversification and structural (utility) demand, and you make the concentrated, high-multiple names prove themselves — or you buy them small, as trades, at a price that pays you for the risk.
Bottom Line
Four record quarters, four falling stocks, one shared fear. The buildout is real and the backlogs are real, but the market is re-pricing how much it will pay for demand that hinges on a handful of hyperscalers. In that environment I'd anchor with EMCOR (EME) for the risk-adjusted core, pair it with Quanta (PWR) as the durable pure-thesis bet on the power bottleneck, treat Sterling (STRL) as a high-conviction rebound trade rather than a foundation, and wait for a better price on Comfort Systems (FIX). The layer of the build with the most defensible demand — the grid — and the name with the best price-to-quality — EMCOR — are where the odds sit today.
Prices and technicals pulled live from Yahoo Finance as of the August 24, 2026 close (10:00 AM HST, market closed). Q2 2026 fundamentals from company releases and BusinessWire/24-7 Wall St./StockTitan/TIKR coverage. Forward P/E figures are approximate, computed against FY2026 guidance midpoints; STRL and PWR are on an adjusted-EPS basis. Levels and multiples are point-in-time and will move.