Why You Should Still Buy The AI Bubble

Analyst reiterates a buy stance on the AI bubble, noting actionable opportunities as ProShare 1X short funds sit at levels typical of bear market lows and forecasting a powerful second wave similar to the internet mania. However, the view is conditional: the team remains cautious around event risk and will only increase equity exposure on confirmation from its short-term ST‑MSI indicator; long position in SPY is disclosed.

Analysis

The next phase of AI upside is unlikely to be uniform: winners will be the firms that convert raw compute into recurring, high-margin monetization (model hosting, add-on enterprise ARR) while losers will be asset-light incumbents that simply relabel features as “AI” without changing unit economics. Expect a two-stage capital cycle — near-term inventory and supply tightness will sustain pricing power for GPUs and memory for 6–12 months, then a second-stage demand wave (endpoint models, verticalized LLMs) drives a multi-year capex cycle benefiting chip designers, memory makers, and toolchain providers.

Key catalysts to watch are concentrated and binary: major datacenter bookings, an outsized GPU guidance beat/miss, or a safety/regulatory event that forces temporary export controls or model throttling. Near-term (days–weeks) volatility will be driven by quarterly prints and options expiries; medium-term (3–12 months) by capex cadence and supply easing; multi-year outcomes hinge on ARPU expansion from software monetization and regulatory regimes that determine cross-border model deployment costs.

Market structure amplifies both upside and downside — concentrated long exposures and thin skew in single-name options create pronounced gamma squeezes on rallies and violent drawdowns on redemptions. That argues for expressed directional exposure via limited-loss option structures and calibrated hedges (portfolio-level put protection or short-dated volatility buys around earnings) rather than naked long equity positions; sizing should assume a 25–40% drawdown scenario in crowded names as a baseline stress case.

Key Decisions for Investors

• Long NVDA via 9–15 month call spread (buy Jan/Feb 2027 1x ITM call, sell 1x further OTM call sized to fund premium). Rationale: captures continued GPU pricing power and second-wave demand with capped premium. Position size: 2–4% NAV; target +50–80% if NVDA sustains >30% move; max loss = premium paid (~100%).
• Buy 3–6 month portfolio tail hedge: SPY 5–7% OTM puts sized to protect 10–15% of equity exposure. Rationale: protects against earnings-driven or macro tightening drawdowns that historically reverse bubbles. Cost: expected drag ~1–3% annualized; payoff asymmetry valuable during 20%+ corrections.
• Relative-value pair: long LRCX (chip equipment) on 15–25% pullback vs short a high-valuation, low-AI-differentiation SaaS name (e.g., CRM-sized position). Timeframe 3–12 months. Rationale: equipment benefits from multi-year capex; short capital light vendors vulnerable to multiple compression if AI fails to boost ARR. Size net exposure to be market neutral ~1–2% net.
• Volatility arbitrage around earnings: sell short-dated OTM calls on hyper-bull names (e.g., NVDA, AMD) after large intraday moves and buy longer-dated puts for asymmetric protection. Rationale: monetize high day-of-event implied vol while retaining long-duration insurance. Keep delta risk small and monitor gamma exposure ahead of prints.