Buy AI-Led EME, PWR and CAT for Long-Term to Gain From Recent Softness

AI data-center infrastructure demand is underpinning a bullish long-term case for EMCOR, Quanta Services and Caterpillar despite recent share-price softness. EMCOR's RPOs reached a record $17.14B, up 43.9% year over year; Quanta's backlog rose 49% to $53.4B; and Caterpillar targets power-generation sales exceeding three times 2024 levels by 2030. Consensus targets imply 26.4%-41.7% upside from recent prices, while all three companies project double-digit revenue and earnings growth supported by data-center power, construction and grid-investment demand.

Analysis

The investable signal is not generic AI exposure but the transition from GPU procurement to constrained physical deployment: interconnection, high-voltage transmission, on-site generation, cooling and commissioning. PWR has the cleanest 6-18 month exposure because grid work is typically earlier in the project cycle and carries a multi-year utility/infrastructure funding tail; EME is more exposed to late-cycle electrical/mechanical fit-out, where labor scarcity can support margins but customer schedule slippage can defer revenue. CAT is the more differentiated power-shortage hedge: distributed generation demand can remain resilient even if individual hyperscale campuses are delayed.

The recent weakness should not automatically be read as an entry signal. These are already consensus AI-infrastructure beneficiaries, and backlog is not equivalent to high-margin, financeable work; the key near-term issue is conversion, not orders. Over the next 1-3 months, earnings commentary on labor productivity, fixed-price project margins, customer deposits and cancellation/deferral rates will matter more than headline backlog. A rise in long rates, utility interconnection delays, or a hyperscaler capex pause would pressure EME/PWR multiples before reported revenue weakens.

Contrarian view: CAT's power-generation optionality is likely underappreciated relative to its cyclical construction-equipment exposure, while EME's apparent discount may appropriately reflect a more concentrated, execution-sensitive data-center mix. Secondary beneficiaries include GE Vernova (GEV) in grid equipment and Vertiv (VRT) in power/cooling, but both offer less valuation protection after AI-driven rerating. NVDA demand remains a necessary but insufficient indicator; power availability, rather than accelerator supply, is increasingly the binding constraint.

Key Decisions for Investors

• Initiate a 6-12 month long PWR / short EME pair at equal dollar exposure. PWR's earlier-cycle grid and utility mix should outperform if data-center builds remain strong but fit-out schedules slip; target 10-15% relative return. Exit if PWR Electric Infrastructure book-to-bill falls below 1.0x or EME demonstrates sustained margin expansion with no project deferrals.
• Accumulate CAT on weakness over the next 1-3 months, sized as a 6-18 month power-constrained AI buildout hedge rather than a broad machinery beta trade. Use a 12-15% downside stop from entry; thesis is falsified if power-generation orders/backlog fail to convert into 2027 guidance or management redirects incremental capacity toward lower-return markets.
• Do not chase EME solely on broker price targets. Set an alert for the next earnings release: add only if RPO growth is accompanied by stable/improving operating margin, favorable cash conversion and explicit confirmation that large data-center customers are not extending construction timelines.
• For broader exposure, favor a modest long GEV versus a short XLI hedge over 3-6 months if transmission/interconnection capex acceleration is confirmed by utilities. The trade benefits from power-grid bottlenecks while reducing general industrial-cycle risk; abandon if regulated-utility capex plans are cut or Treasury yields reprice materially higher.