Powell Industries

Powell Industries is the smallest pure-play in the AI-data-center electrical-infrastructure trade — and the highest-leverage one. While Eaton, Schneider and Vertiv dominate headlines, Powell builds the medium-voltage switchgear and bus duct assemblies that connect data-center substations to the row-level distribution. FY2025 revenue of $1.21 bn (+33% YoY) hides the actual story: backlog at fiscal year-end stood at $2.62 bn — more than 24 months of revenue at current run-rate, the highest book-to-bill in the company's 80-year history.

Two structural drivers explain the orbital trajectory. First, hyperscaler data-center capex is breaking historical patterns — AWS, Microsoft, Google and Meta combined are guiding to over $310 bn of FY2026 capex (consensus), with 20-25% directed at electrical infrastructure. Second, the LNG-export buildout in the Texas Gulf Coast — Powell's traditional core market — added a new layer of capacity orders in H2/2025 from Sempra, Cheniere and Venture Global FID approvals.

The 2026 question is execution. Powell has been capacity-constrained since Q2/2024 — backlog grew faster than ability to fabricate. The Q4/2025 expansion of the Houston North campus added 30% manufacturing capacity, with second-line ramp scheduled Q3/2026. If execution delivers, FY2027 revenue could hit $1.8-2.0 bn at gross margin above 28% — versus 25% in FY2025.

Top holders Q1/2026: BlackRock 12.4%, Vanguard 11.1%, Royce & Associates 4.8%, Dimensional 4.5%. The notable name is Royce — a small-cap quality specialist whose ownership signals an exit-rated quality screen. No insider-driven hedge funds (no Coatue, Tiger Global) have meaningful positions.

What is interesting: Bill Ackman discussed Powell publicly at the Ira Sohn 2026 conference in May as a long position — not as a major holding (Pershing Square has not yet filed) but as a thesis-illustrative AI-electrification trade. Sohn coverage triggered a 12% single-day rally.

Insider activity: CEO Brett Cope sold $4.2 M of stock in November 2025 at $260 (now $292, +12% since) under 10b5-1, but the sale represented only 8% of his holdings — viewed as routine. CFO Michael Metcalf bought $180k of shares in February 2026 at $245.

Short interest at 6.3% — relatively elevated, reflecting cyclical-industrial late-cycle fears. Days-to-cover sits at 4.1 days.

Forward P/E 42.6x vs industrial-cyclical median 21x — at the upper end of any historical range. EV/Sales 4.2x vs the same peer group at 2.1x. The premium reflects the backlog visibility and AI-data-center mix shift. EV/EBITDA 19.4x at the lower end of US AI-infrastructure peer set (Vertiv 23x, Eaton 17x). Sell-side PT range $240-$385 (median $315): Wells Fargo most bullish at $385 (FY2027 EPS $9.20 + 28x multiple), Morgan Stanley most bearish at $240 (assumes capacity expansion 6-month delay). Implied probability of execution success in current price is roughly 55%. Bull case $420 if FY2027 revenue hits $2.0 bn at 30% gross margin. Bear case $180 (-38%) if hyperscaler capex moderates and backlog cancellations begin.

1. May 2026: FY-Q2/2026 earnings — first quarter with new capacity online, gross-margin trajectory critical
2. Q3 2026: Houston North second-line capacity ramp complete — execution proof point
3. Microsoft FY27 capex guidance (calendar July 2026): Single largest AI-capex datapoint that drives Powell sentiment

Powell is the high-conviction high-volatility leveraged play on AI-data-center electrification. The backlog is real and the capacity expansion is on track — but at 42.6x forward, the market is pricing near-flawless execution. I find the asymmetry on the small side: maybe 30% upside in a clean execution scenario, but 30-40% downside if Houston North delivers late or hyperscaler capex moderates. I size POWL at 1.5-2% because of the execution-dependent margin story plus the high beta to single AI-capex headlines. Trigger to add aggressively: any Q2/2026 print showing gross margin expansion above 27% AND fresh backlog growth above $2.8 bn. The trade I would not make is going underweight — small-cap AI-infrastructure has too much momentum to short into.