Forget Chips: These 3 Stocks Are Building the AI Data Center Boom

JPMorgan estimates ~60% of planned data-center capacity for 2027 hasn’t even broken ground yet, helping explain 2026 whipsaws in AI-linked stocks—but also highlighting upside for construction bottleneck beneficiaries. Comfort Systems (FIX) reported a record $14.1B backlog (+73% YoY) and nearly $1.0B in FCF, with consensus expecting ~20% upside; EMCOR (EME) posted Q2 revenue of $5.15B (+19.8% YoY), +35% YoY adjusted EPS, and raised guidance (consensus ~19% upside). Sterling Infrastructure (STRL) delivered a Q2 revenue +90% YoY to $1.17B and adjusted EPS +116% YoY to $5.80, with combined backlog $5.62B (+150% YoY) and full-year guidance raised, supporting further upside despite valuation pullback.

Analysis

The market is likely still pricing this as a generic AI capex theme, but the real monetization layer is labor-constrained project execution. That shifts margin power from chip vendors to contractors with scheduling control, procurement scale, and balance-sheet flexibility; among the names here, EME looks best positioned to compound because its mix is less dependent on a single end-market, while STRL has the highest operating leverage if starts keep accelerating. FIX is the purest backlog story, but it is also the most vulnerable to any slowdown in new awards because its valuation already assumes near-flawless conversion.

The key risk is that backlog is not the same as billings. If utilities, permitting, or trade labor bottlenecks persist, revenue timing can slip even when demand remains intact, which would hit multiples before fundamentals show up in reported numbers. Over the next 1-3 months, watch for any evidence that data-center starts are being deferred rather than merely delayed; over 6-18 months, the secular winner is whoever can turn scarce electricians/plumbers/welders into repeatable capacity, not just land the order.

Consensus seems to be extrapolating linear backlog conversion into linear earnings upgrades, but that is probably too clean. The contrarian setup is that the first leg of upside may already be in the shares, while the second leg depends on margin durability and working-capital discipline, not just top-line growth. If labor inflation spikes or backlog growth decelerates below revenue growth, the group can de-rate quickly even if the AI buildout story remains intact.

Key Decisions for Investors

• Long EME / short FIX as a 3-6 month relative-value pair: favor EME’s lower multiple and broader earnings base over FIX’s higher-multiple, single-theme exposure. Target 10-15% relative outperformance for EME; thesis breaks if FIX continues to outgrow backlog conversion and sustains premium margin expansion.
• Buy STRL only on material pullbacks or via call spreads into earnings over a 6-12 month horizon. It has the best operating leverage to data-center execution, but the 37x P/E leaves little room for a miss; invalidation is any guidance reset or backlog growth slowing meaningfully below revenue growth.
• Use FIX as a momentum trade, not a core valuation buy: if entering, prefer partial size or a call spread with a 3-6 month horizon. Risk/reward is favorable only if the next earnings cycle confirms that backlog is converting without margin compression.
• Set a watch item on labor-cost indicators and project-start data for the next 1-2 quarters; if construction wage inflation accelerates or starts are delayed, trim all three names because the market will likely compress multiples before fundamentals roll over.