Powering AI: Top Engineering, Procurement, and Construction Stocks to Buy

Barclays says annual AI infrastructure spending by Western hyperscalers and AI labs could exceed $1 trillion, more than $300 billion above current consensus, with peak spending expected in 2028. The firm highlighted EPC and infrastructure names such as Quanta, EMCOR, MasTec, AECOM, and Fluor as key beneficiaries of the data center and power buildout. The note is constructive for the group, but it is an analyst thematic call rather than a company-specific catalyst.

Analysis

The market is still underestimating how non-linear EPC demand becomes once hyperscalers and utilities are forced into the same queue for transformers, switchgear, generators, and skilled labor. The near-term winners are the names with the tightest control over scarce execution capacity and the highest mix of power-adjacent work, because pricing and backlog duration tend to expand before revenue does; that favors PWR, EME, STRL, and MTZ over more generalized engineering shops. The second-order beneficiary is the electrical and low-voltage supply chain, where lead times remain the real bottleneck, so vendors with exposure to conduit, cabling, and controls can still enjoy margin lift even if headline data center spend gets revised down later.

The setup is not just about AI capex growth; it is about capex persistence. If training demand decelerates by the back half of the decade, the overhang shifts from compute racks to power delivery, campus completion, and grid interconnect work, which extends the earnings runway for contractors even after the model-building euphoria fades. That makes the cycle look more like a multi-year infrastructure build than a one-off technology surge, and it reduces the risk that a temporary AI digestion phase immediately hits the group.

The main risk is valuation compression before fundamentals catch up. These stocks have already started to discount a durable backlog upcycle, so any signs of customer capex rationing, grid permitting delays, or labor inflation could hit multiple expansion faster than consensus expects. The most vulnerable names are the more cyclical or less differentiated contractors where margin quality is weaker and order visibility is lower, especially if investors rotate toward cleaner ways to play the theme via utilities or power equipment instead.

Contrarianly, the consensus may be too focused on who installs the boxes and not enough on who monetizes the power constraint. The better risk/reward may be in firms that can capture the highest-value scopes per project rather than the broadest revenue growth, because scarcity economics usually accrue to specialized execution rather than diversified share counts. In that frame, the market likely still has room to rerate the highest-quality execution names, while lower-conviction participants are more likely to become source of alpha for shorts or underweights.

Key Decisions for Investors

• Long PWR / short a basket of lower-quality contractors (e.g., TPC, MYRG) for 3-6 months: capture scarcity premium in power interconnect work while avoiding names with weaker pricing power and execution quality.
• Initiate a tactical long in STRL on pullbacks over the next 1-2 weeks: strongest operating leverage to early-phase site development, with favorable upside if backlog converts faster than the market expects; trim if capex headlines decelerate.
• Overweight EME versus ACM for 6-12 months: EME has cleaner mission-critical electrical/HVAC exposure and should reprice more quickly as data center scope shifts toward power density rather than design services.
• Use FLR as a higher-beta expression only if you want optionality on adjacent energy and nuclear work; otherwise keep it smaller versus PWR/EME because the catalyst set is broader but less pure, implying lower certainty of multiple expansion.
• Fade KBR and MYRG relative to the core winners: lower signal from the article and weaker thematic leverage means they are more exposed if the market starts distinguishing true AI infrastructure beneficiaries from generic EPC exposure.