Worried about AI? Become a plumber, says Nvidia’s billionaire boss

Tech firms plan to spend $660bn on AI data centres this year (up 60% y/y), supporting what Nvidia CEO Jensen Huang calls a multi-trillion-dollar infrastructure build-out. Nvidia, now valued at $4.4tn, positions it as a primary beneficiary while Huang highlights demand for skilled manual roles (electricians, plumbers, technicians) rather than mass job losses. Risks include Oracle’s recent job cuts, scrapped expansion with OpenAI, and rising energy costs/consumer electricity concerns that could constrain data-centre expansion despite a White House pledge by tech firms to help fund increased electricity output.

Analysis

A concentrated wave of infrastructure capex will redistribute economic value down the stack: equipment OEMs, power-grid and transformer suppliers, specialty materials (copper, specialty gases) and large EPC contractors are poised to capture recurring, higher-margin revenue that hyperscalers and chip designers cannot fully internalize. Expect margin pools to shift toward firms that own physical installation and grid augmentation capabilities; this will create durable orderbooks for suppliers but also raise input-cost inflation for cloud operators over a 12–36 month horizon.

Key risks that could unwind the trade are macro-driven capex pauses, a sustained spike in electricity prices, or rapid regulatory limits on buildouts in constrained grid regions. Near-term catalysts to watch: quarterly capex guidance from hyperscalers (next 1–3 quarters), regional transmission permitting outcomes (3–18 months), and any acceleration of export controls or tariff regimes that materially alter fabs’ geography.

The consensus underappreciates two second-order effects: (1) labor tightness and wage inflation in manual roles will compress returns to owners of compute if contractors capture more of the ROI, and (2) rising operational energy costs create a nonlinear margin squeeze that could make some planned projects uneconomic if power prices remain elevated. This creates a bifurcated opportunity set—own physical-infrastructure capture vs. selectively hedge software/cloud exposure where energy risk is concentrated.

Key Decisions for Investors

• Long NVDA (core position): initiate a 6–12 month call spread sized 3–5% of tech book (buy 9–12 month ITM calls / sell higher strike calls). Target +50–80% if enterprise capex guidance accelerates; cut to flat on -30% relative move or if hyperscaler guidance shows sequential capex cuts.
• Pair trade — long AMAT or LRCX (capital equipment exposure) / short ORCL (enterprise software execution risk): allocate 2–4% notional long AMAT or LRCX via 9–12 month call buying, financed by shorting ORCL stock or buying 12-month 10% OTM puts. Expect asymmetric payoff: 30–60% upside in equipment names vs limited payout if ORCL negative headlines persist; stop-pair if spread narrows <15% from entry within 6 months.
• Infrastructure/Power tilt — long ETN (Eaton) or ABB via 12-month calls: size 2% to play grid augmentation demand and transformer orders. Target 25–40% return as utilities and hyperscalers spend on distribution upgrades; downside limited to premium paid (~100% loss of option premium) if projects stall due to permitting.
• Hedge the macro — buy 6–9 month puts on AMZN and MSFT (small hedge ~1–2% of portfolio) to protect against a capex shock that would compress multiple expansion in cloud-exposed names. If capex stays strong, expect these hedges to expire worthless; they pay off >2x notional if broad capex guidance falls by >15% over two quarters.