Wall Street is pricing ACA — and the entire utility structures sector — like the datacenter infrastructure boom is peaking. They're wrong by about three years.

The Lag Mechanism

ACA's CEO dropped the key insight unprompted on the Q4 2025 call: "We do not sell to a hyperscaler. Our customers are the people who supply power to developers and hyperscalers. It might take two to three years for us to start seeing any noise."

Read that again. ACA makes the poles and structures that utilities need to connect power to datacenters. Their customers are Duke Energy, Southern Company, NextEra — not Microsoft or Amazon. And there's a 2-3 year cascade between a hyperscaler breaking ground and ACA getting an order for the transmission infrastructure.

This means ACA's current healthy demand — utility structures backlog $435M, volumes up double digits, pricing up high single digits, sequential margin expansion every quarter of 2025 — reflects datacenter investment decisions from 2022-2023. The pre-AI wave.

The 2024-2026 AI capex surge? The $300B+ that AMZN/MSFT/GOOG/META are pouring into compute? That hasn't hit ACA's order book yet.

Eight Companies, Zero Contradictions

We checked this against every relevant peer. The result was unanimous:

Eaton (ETN) quantified the lag directly: AI is 50% of new orders but only 30% of current revenue. That 20-point gap IS the lag in dollar terms. Mega projects convert to revenue "over three to five years." Their datacenter backlog represents 11 years of construction at 2025 build rates. Start rate on tracked mega-projects: 16%. Meaning 84% of the pipeline hasn't broken ground.

Duke Energy (DUK) named the inflection year: ESAs signed today "start coming online late 2027, ramp 2028." They raised their 5-year capital plan to $103B (+18%). Data centers went from 50% to 75% of their load growth pipeline in a few quarters.

Southern Company (SO) confirmed the same timing: revenue growth "into 2027, more pronounced expansion 2028." They raised their plan to $81B (+30% from a year ago). 10 GW signed, another 10 GW in late-stage discussions. Annual electricity sales growth projected at 10% through 2030 — raised by 2 full percentage points.

VMI (closest ACA peer in utility poles): backlog $1.5B (+22% YoY), spending $170-200M on capacity expansion in 2026. NUE is building 4 greenfield utility pole plants from scratch — you don't do that unless you see a decade of demand. Hubbell reported Q4 "acceleration versus prior quarters" — explicitly. Quanta Services hit $44B record backlog and is spending $500-700M to build its own transformer factory because the supply chain is too constrained.

Eight companies. Three continents of operations. Zero said demand is flattening. Every single one reported acceleration.

The Capacity Shortfall Is Real

ACA commissioned a third-party study that confirmed a 2027 capacity shortfall in utility structures. They then announced a second wind tower facility (Tulsa, Oklahoma) converting to utility pole production — making this a systematic playbook, not a one-off.

The conversion economics are elegant: ACA takes stranded wind tower manufacturing capacity (policy headwinds from OBBBA killing IRA credits) and redeploys the same people and roughly the same equipment to make utility poles. Lower capex than greenfield, faster time to market, and people are the binding constraint per CEO Carrillo.

Illinois (Clinton) comes online H2 2026. Tulsa transitions as wind orders complete through 2027. Meanwhile VMI is expanding aggressively and Nucor is entering from scratch with 4 plants. The market can apparently absorb all of this simultaneously.

The Vehicle Problem

Here's where I have to be honest about ACA specifically.

Factor regression: 65.6% idiosyncratic variance. Below our 75% threshold. XLI (broad industrials) explains 23% of ACA's return variance. Market beta explains another 15%. The stock moves with the sector.

The conversion playbook is genuinely ACA-specific — maybe 20% of variance. But the strongest evidence we found (the lag mechanism, the capacity shortfall, the demand acceleration) supports the SECTOR factor, not the company. VMI, Hubbell, Eaton, Powell — they all ride the same wave.

Comparison:

POWL is more idiosyncratic but you're paying 32x for a stock that's already tripled. ACA is cheaper but less differentiated. Neither clears 75% cleanly.

What's Actually Tradeable

The sector insight is the alpha, not the stock pick. Three ways to express it:

Single stock (if you must): ACA at RSI 29 and 20x forward earnings is the value entry. The conversion playbook creates some idio. Forward alpha ≈12% EV-weighted.

Pair/basket: Equal-weight ACA/VMI/POWL/HUBB captures the sector thesis without single-stock risk. This is probably the right answer.

The contrarian play: POWL at RSI 22 with 16.8% short interest and its first datacenter mega-project. If datacenter orders accelerate, shorts get squeezed. Higher risk, higher idio, much richer valuation.

Risks That Matter

Mexico tariffs. ACA has 1,225 employees manufacturing utility structures and wind towers in Mexico. USMCA first review is July 2026. Management didn't mention it on the Q4 call — notable silence. This is the underappreciated risk.

Wind tower drag. 2026 revenue stepping down 25%. Margins guided "flat at best" for the combined Engineered Structures segment. The conversion thesis assumes utility growth offsets wind decline — management is cautious about guaranteeing that.

Capacity overshoot. ACA converting 2 facilities, VMI spending $170-200M, NUE building 4 greenfield plants — all simultaneously. If the demand wave disappoints or arrives late, there's a 2028-2029 glut risk.

The Number That Matters

ETN tracks 866 mega-projects worth $3 trillion. Start rate: 16%. The other 84% is coming. Grid infrastructure — the poles, the transformers, the substations, the switchgear — must be built before a single watt flows to a single GPU.

The market is pricing this sector like we're at the top. The lag says we're in the third inning.