AI Spending Frenzy Echos Early Go-Go Days of Shale, Currie Says

Carlyle veteran Jeff Currie warns in a research note that Big Tech’s heavy spending on artificial intelligence mirrors the profligate investment era of shale, which preceded a price collapse that erased roughly $2.6 trillion in equity. Currie argues that energy and technology together underpin the broader economy, implying that excessive capital deployment in AI could create systemic vulnerability and weigh on investor positioning across other sectors such as finance and health care.

Analysis

Market structure: The immediate beneficiaries are GPU designers and fabricators (NVDA, AMD), lithography/equipment (ASML), hyperscale cloud providers (MSFT, GOOG, AMZN) and data‑centre REITs (EQIX); losers are capital‑starved small‑cap AI pure‑plays and legacy enterprise software with weak monetization. Pricing power is consolidating—Nvidia‑style architectures and cloud scale create winner‑take‑most dynamics that can sustain >20–30% gross margins for leaders while compressing smaller vendors. Tight GPU wafer/packaging supply implies continued backlog for 6–12 months but risk of overcapacity if capex accelerates, pushing oversupply in 12–24 months and downward price pressure on chips and power demand volatility.

Risk assessment: Tail risks include export controls/antitrust actions (weeks–months) that could cut China revenue by >10–20%, a rapid writedown cycle from overbuilt AI infrastructure (quarterly), and energy/grid shocks that lift data‑centre opex >10% year‑over‑year. Immediate moves (days) will be momentum-driven by guidance; short term (months) dependent on order flow and component lead times; long term (3–36 months) depends on software monetization and model‑run economics. Hidden dependencies: ML talent inflation, rare‑earth/wafer bottlenecks, and local power capacity; catalysts to watch: capex guidance, large cluster orders, semiconductor export policy, and utility rate filings.

Trade implications: Tactical: allocate 1–2% notional to NVDA via 3–6 month call spreads (limit downside; add on 10–15% pullbacks) and 2–3% long in ASML for 12–24 months to play equipment lead times. Buy 1–2% exposure to EQIX or Digital Realty for structural demand and sell covered calls to harvest income. Short 1–2% in basket of small‑cap AI pure‑plays (market cap < $3bn) via 3–6 month put spreads where valuation >20x revenue; size modestly and scale on earnings misses. Hedge systemic risk with 0.5–1% SPX 5–7% OTM puts if negative headlines or GPU price drops >25% in 90 days.

Contrarian angles: Consensus underestimates front‑loaded capex and second‑order margin pressure from talent and power costs—this creates a 12–24 month oversupply risk that can compress multiples by 20–40% for non‑dominant players. The shale analogy is informative on boom/bust but AI has stronger network effects and SaaS monetization, so leaders may avoid shale‑style wipeouts; the market may be overpricing perpetual growth for peripheral AI names. Watch for two technical thresholds as early signals: GPU spot prices down >25% vs three months ago (oversupply) and hyperscaler capex guidance cut by >10% year‑over‑year (demand shock).

Key Decisions for Investors

• Establish a 1–2% portfolio position long NVDA via a 3–6 month call spread (buy ATM call, sell 30–40% OTM) to capture upside while capping cost; add another 0.5–1% on any 10–15% pullback; target 6–12 month horizon.
• Initiate 2–3% long in ASML (equipment exposure) and 1–2% long in EQIX or Digital Realty for 12–24 months; trim ASML if it appreciates >25% or EQIX >20% or if hyperscaler capex guides down by >10% YoY.
• Short 1–2% combined exposure to small‑cap AI pure‑plays (market cap < $3bn) via 3–6 month put spreads, focusing on names trading >20x revenue with negative free cash flow; set stop‑loss at 15% adverse move and tighten if borrow costs spike.
• Purchase a 0.5–1% portfolio tail hedge with 3–6 month SPX 5–7% OTM puts if either: (A) GPU spot prices decline >25% versus 90 days ago, or (B) significant export‑control headlines or hyperscaler capex cuts >10% YoY appear within 30 days.