Big Tech Will Spend $735 Billion on Data Centers. This Boring Electrical Contractor Collects the Checks.

Quanta Services reported first-half 2026 revenue of $17.4 billion, up 34% year over year, while net income rose 80% to $672 million; its backlog reached a record $53 billion, with 60% expected to convert to revenue within 12 months. Driven by hyperscaler AI data-center investment and demand for power, cooling, grid interconnection, and skilled labor, Quanta raised 2026 guidance to roughly $39.5 billion in revenue, $2.0 billion-$2.5 billion in free cash flow, and midpoint EPS of $16.70, up 55% from last year. The company is positioned as an infrastructure beneficiary of an estimated $735 billion in annual data-center spending by Amazon, Microsoft, Alphabet, and Meta.

Analysis

PWR’s differentiated exposure is not merely data-center construction volume; it monetizes the permitting, interconnection, transmission, and field-labor bottlenecks that determine whether announced campuses become operating compute capacity. This should support mix and margin resilience versus more commoditized electrical contractors, while creating read-through upside for grid-equipment suppliers ETN, HUBB and GEV. The second-order constraint is utility queue duration: if power-delivery schedules, rather than building shells, become the gating item, hyperscalers may shift toward regions with available generation or fund dedicated generation—both outcomes expand off-site transmission and substation work.

The near-term risk is that PWR already reflects a widely recognized AI-infrastructure narrative, leaving little tolerance for margin dilution from accelerated hiring, subcontractor costs, or fixed-price project execution. Backlog is not equivalent to earnings quality: investors should focus on book-to-bill, backlog gross-margin mix, customer concentration, cancellation provisions, and working-capital conversion over the next two quarters. A slower hyperscaler capex cadence would initially hurt suppliers with discretionary data-center exposure such as VRT more than PWR, but a sustained pause would ultimately compress PWR’s growth multiple.

Consensus may be underestimating the duration of grid modernization demand independent of AI, which gives PWR a better downside floor than pure data-center plays. Conversely, it may be overestimating the speed at which announced AI campuses convert to revenue; transmission approvals and utility interconnection can defer projects by years. The key 6-18 month catalyst is evidence that large campus awards are converting into higher-margin power and mechanical work rather than lower-margin general EPC scope.

Key Decisions for Investors

• Maintain a core long PWR only on pullbacks or following quarterly confirmation of stable/improving segment margins and cash conversion; target a 6-12 month holding period. Do not chase a multiple expansion without updated backlog-margin disclosure.
• Use a relative-value basket: long PWR and ETN versus short VRT in a modestly sized 3-6 month pair if data-center power constraints intensify. PWR/ETN have broader grid and utility demand support; VRT is more exposed to a near-term pause in AI campus fit-outs.
• Set a downside thesis stop if PWR reports two consecutive quarters of declining book-to-bill, material backlog conversion slippage, or free-cash-flow conversion below management’s implied run rate; these would indicate labor and project-execution pressure rather than temporary timing.
• Watch utility interconnection timelines and hyperscaler capex guidance at AMZN, MSFT, GOOG, and META. A coordinated capex deferral or evidence of campus cancellations is a signal to reduce PWR before reported backlog weakens, typically with a 1-2 quarter lag.