The Real Money in AI Might Be in Power Cooling and Connectivity

AI infrastructure 'pick-and-shovel' names are outpacing marquee AI software companies on recent growth metrics: over the last three years trailing-12-month revenue grew 92.8% at Arista and 70.4% at Vertiv versus 44% at Microsoft and 37.3% at Alphabet. Net income gains are even starker — Arista up ~148.2% and Vertiv up ~1,250% (though absolute net incomes remain much smaller: roughly $3.4B for Arista and $1B for Vertiv versus $100B+ for the tech giants). The market has priced that out: forward P/Es are roughly 45.8x for Arista and 40.6x for Vertiv versus ~30x for Microsoft and Alphabet, signaling investors have already baked-in higher growth expectations even as questions about AI spending sustainability and competitive dynamics (OpenAI, Anthropic, Google) persist.

Analysis

Market structure: Data‑center infrastructure (Arista ANET, Vertiv VRT, switch/thermal suppliers) are clear near‑term winners as cloud/AI capex shifts from software to buildout — ANET revenue +92.8% and VRT +70.4% TTM three‑year growth. The market already prices this (ANET fwd P/E ~45.8, VRT ~40.6 vs MSFT/GOOGL ~30), implying growth expectations baked into multiples and limited margin for execution misses.

Risk assessment: Key tail risks are a sudden cloud/AI capex slowdown (≥10% QoQ downgrade), concentration risk (loss of one hyperscaler customer could cut revenue by >15%), and higher rates compressing growth multiples; Vertiv’s 1,250% net‑income jump looks idiosyncratic and vulnerable to reversion. Immediate signals (days–weeks): earnings guidance and hyperscaler capex commentary; medium (3–12 months): order backlogs and channel inventory; long (12–36 months): secular AI infrastructure adoption vs commoditization.

Trade implications: Favor overweight to infrastructure suppliers but size conviction: prefer ANET for quality growth, VRT as higher‑beta exposure. Use pair trades (long ANET / short GOOGL or MSFT) to isolate infra vs platform beta. Options: use 9–18 month LEAPS to leverage growth and 3–6 month puts for downside protection around earnings cadence.

Contrarian angles: Consensus overlooks margin cyclicality and contract concentration — if hyperscalers internalize more networking or OEM pricing pressure rises, ANET/VRT multiples can compress >20% quickly. Conversely, a second wave of AI capex (new model launches) could make current premiums look cheap; watch cloud capex revisions and inventory‑to‑sales ratios for tipping points.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in Arista Networks (ANET). Action: buy into weakness up to a 15% intraday pullback or if forward P/E compresses to ≤38. Target hold 12–24 months, price target +30–60%; place a hard stop at 18% below entry or exit if two consecutive quarters miss revenue guidance by >5%.
• Allocate 1–2% to Vertiv (VRT) as a tactical, higher‑beta AI‑infrastructure exposure. Preferred execution: buy shares on a >20% pullback or purchase Jan 2028 LEAPS ~15–25% OTM to cap downside. Take profits if revenue growth slows to <15% YoY in two sequential quarters or if gross margin contracts >300bps.
• Implement a pair trade: long ANET (2% of portfolio) / short GOOGL (1.5%). Rationale: capture valuation spread and infra outperformance; target 10–20% relative outperformance in 6–12 months. Cover/close if GOOGL reports cloud capex guidance >+10% YoY or if ANET stock rises >40% from entry.
• Use options to manage risk and harvest yield: sell 3–6 month covered calls on MSFT or GOOGL positions to generate 2–3% quarterly yield if held, and buy 9–12 month protective puts 5–10% OTM on new ANET/VRT longs sized to cover ~50% of position cost (hedge earnings/visibility risk).