Ray Dalio says AI is in ‘the early stages of a bubble,’ so watch out for 2026

Bridgewater founder Ray Dalio warns that the AI-driven surge in tech stocks looks bubble-like and could face a correction in 2026, estimating current euphoria at roughly 80% of historic bubbles. He flags Fed policy uncertainty—Powell's term ends in May and a Trump-appointed successor favoring lower rates could further inflate valuations—while noting market distortions last year: S&P 500 up 16%, MSCI Emerging Markets up 33%, gold was the top major market and outperformed the S&P by 47%, and the U.S. dollar fell about 10%; Dalio urges diversification given these risks.

Analysis

Market structure: The primary winners in a late-stage AI euphoria are capital-intensive providers of compute and software platforms (NVDA, TSM, MSFT, GOOGL) and commodity/FX beneficiaries like gold (GLD/GDX) and EM exporters; losers are high-multiple, early-stage AI adopters that haven’t proven unit-economics (small-cap SaaS, unprofitable cloud integrators). A weaker dollar and lower rates mechanically reprice global equities (EM and Europe up) and commodities up; concentrated index flows to mega-caps raise single-name systemic risk and skew implied vol curves higher for tech. Supply/demand: GPU and HBM supply constraints sustain pricing power for chip suppliers through 2026 unless capex ramps 50%+; gold demand is bid by fear flows and jewelry/central-bank buys, tightening available investable supply.

Risk assessment: Tail risks include a regulatory crackdown on AI models or large-scale model failures (data breaches, safety incidents) causing >30% drawdowns in affected names, or a U.S. Fed policy surprise (hawkish) that strengthens the dollar and shocks EM and commodity prices. Immediate (days) risks center on volatility around Fed-replacement signals and quarterly earnings; short-term (0–6 months) risks include profit-cycle disappointments from AI pilots (MIT 95% pilot failure metric) and index rebalancings; long-term (1–3 years) risks are structural: AI integration failing to lift margins widely. Hidden dependencies: valuations masked by a -10% dollar; buybacks funded by cheap debt; concentration in a handful of ETFs.

Trade implications: Reduce concentration in top AI longs while selectively keeping exposure to structural winners; hedge with puts rather than outright shorts to limit convex risk. Favor long GLD/GDX and selective EM exposure (EEM, TSM) if dollar remains weak; prepare to increase duration (TLT) on dovish Fed confirmation. Use options to buy downside protection on NVDA/QQQ and volatility on SPY around the May 2026 Fed appointment.

Contrarian angles: The consensus of an imminent AI crash misses durable moats: NVDA’s datacenter pricing power and TSM’s foundry scarcity may continue to compound cash flows even if broader AI froth corrects. Overdone trades: blanket short tech/QQQ is risky — prefer pair trades and hedged option structures; underdone: long selective miners and high-quality EM exporters with pricing power. Historical parallels (2000 tech vs 2008 commodity cycles) show bubbles can coexist with structurally rising commodities; prepare for asymmetric outcomes.

Key Decisions for Investors

• Trim concentrated mega-cap AI positions (NVDA, MSFT, GOOGL, META) by 30–40% of current weight over next 2–6 weeks; replace size with 1–2% portfolio hedges by buying 3-month ATM put spreads on NVDA and QQQ sized to cover the trimmed notional (cap cost to ~0.5–1% portfolio).
• Establish a 3–5% portfolio allocation to gold: 60% GLD, 40% GDX immediately as crisis/FX hedge; take 50% profits if GLD rallies +15% from today's levels or if DXY strengthens +5% within 60 days.
• Implement a pair trade: long EEM 2.5% vs short QQQ 1.25% (net long risk to EM) to capture potential re-rating if dollar remains down into May 2026; rebalance after Fed appointment or if DXY moves >±7% from current level.
• Buy a 0.5–1.0% portfolio-sized 3-month ATM straddle on SPY ahead of the Fed chair decision window (April–May 2026) to hedge systemic volatility; sell if implied vol decays relative to realized by >20% or after Fed decision.
• Increase duration exposure with a tactical 2–3% allocation to TLT if 10-year Treasury yield falls >25 bps from current levels or after a dovish Fed successor is confirmed; cut to neutral if 10-year yield rises above a 50 bps move from entry.