One AI bubble has already burst. The next one—a ‘rare’ kind—is still growing, economist warns

Capital Economics' John Higgins concludes the AI valuation bubble has burst as IT/Big Tech price-earnings ratios have fallen to their smallest level since the pandemic (after peaking near ~75% in late 2024). There are 498 AI unicorns valued at $2.7 trillion and OpenAI was reported at $730B (up from $500B six months earlier); Salesforce and ServiceNow are each down ~30% YTD, while Nvidia reported Q4 revenue of $68.1B (+73% YoY). Higgins warns the greater risk is an 'earnings bubble' — Goldman Sachs projects $539B in AI capex for 2026 — and macro/geopolitical shocks (e.g., Iran war cutting Qatar helium, ~1/3 of global supply) could trigger a sharp earnings and sector correction.

Analysis

Price action has likely removed a lot of the easy upside in AI-exposed tech, but the more dangerous second-order risk is that reported earnings themselves are a cyclical bubble. Many AI winners exhibit high operating leverage and front-loaded R&D/capex: if enterprise adoption stalls or macro demand softens, revenue growth can flip to mid-single-digit declines within 2-4 quarters while cost bases lag, producing outsized EPS compression relative to the top-line move.

SaaS incumbents with subscription models and multi-year renewals face asymmetric downside from two directions — multiple compression driven by investor sentiment and a structural churn risk as customers reallocate spend to cloud infra, system integrators and specialist AI tooling. Semiconductor and materials suppliers sit on a fragile intersection of demand front-loading and concentrated commodity constraints; a localized supply shock (helium/energy or logistics) can tighten gross margins across the stack for 1-3 quarters even if end-demand remains intact.

Key catalysts to watch: 1) corporate capex schedules and 2) quarterly renewal/attrition metrics from large enterprise customers — these will show whether AI is additive or simply reallocating existing IT budgets. Tail risks are short-dated and event-driven (geopolitics, supply interruptions) with 0–6 month impact, while earnings mean-reversion from adoption fatigue plays out over 6–18 months. Monitor dispersion between implied vol and realized vol: options markets are pricing asymmetric outcomes into a handful of names, creating tactical entry windows.

Contrarian angle: sentiment may be over-penalizing generational platform owners while underweighting the durable service revenue that will glue ecosystems together (cloud infra, managed services, chip foundry partners). A targeted, hedged long of select infrastructure exposures versus short-duration SaaS beta captures that skew with defined downside protection.

Key Decisions for Investors

• Initiate a paired trade: long NVDA via a 6–12 month call spread to capture continued AI infra demand (defined-cost, 2–4x upside if growth persists) funded by selling 3–6 month covered calls or call spreads on CRM (collect premium vs. downside risk in SaaS). Rationale: capture asymmetry between infrastructure winners and vulnerable SaaS multiples; stop-loss if NVDA falls >20% on supportive volume or CRM rallies >15% on news.
• Buy 3–6 month put spreads on CRM and NOW (e.g., a 20–40% width) to express downside from subscription churn and budget reallocation; max loss = premium paid, target payoff if names decline 25–40% within the window. Size these positions small (2–4% net portfolio) given event and liquidity risk.
• Deploy a volatility pair: buy near-term NVDA straddle/long call skew ahead of its next major guidance event (30–60 days) while simultaneously selling short-dated call premium on CRM/NOW to finance cost. This captures asymmetric NVDA upside or guidance beat while monetizing elevated SaaS call demand; be mindful of IV crush on NVDA after earnings—use spreads to limit Vega exposure.
• Use GS flow as a signal, not a core position: scale a tactical long of GS (2–3% sleeve, 3–6 month horizon) ahead of macro/capex reports — if GS upgrades AI capex outlook, buy-write to harvest calm premium; if GS downgrades, use proceeds to add to SaaS puts or NVDA hedges.