The economist who said ‘no bubble’ just sounded the alarm: ‘The season of chaos is at hand’

Return dispersion in the world stock market has surged to near dot-com-era extremes, with April 2026 dispersion at 15.0% and May at 15.6%, versus a 7.6% average since 1995. Micron rose 87.8% and SK Hynix 78.6% in May, together contributing 17% of the MSCI ACWI’s 5.2% monthly return despite weights below 1%. The article argues that AI-driven exuberance, record IPO activity, and elevated earnings expectations are reviving bubble concerns and keeping volatility high.

Analysis

The market is transitioning from a broad-index regime to a narrow-leadership regime where a few AI-adjacent winners dominate returns. That is usually a bad environment for passive beta and mediocre for diversified long-only books, because index-level performance can look fine while breadth deteriorates underneath; active underperformance becomes a mechanical issue, not a thesis issue. The immediate second-order effect is that capital formation for late-stage AI infrastructure may be getting easier just as scrutiny on monetization gets harder, which tends to pull multiple groups higher before fundamentals can validate them.

The more important signal is not valuation alone but the implied fragility of earnings expectations. When forecast growth is pushed far above realized history, the market is effectively pricing perfection across a long horizon; that makes the tape highly sensitive to any capex delay, customer concentration issue, or margin guide-down from the AI supply chain. In that setup, hyperscaler beneficiaries can keep working for months, but the downside tail becomes convex once one or two bellwethers show that spending is being deferred rather than merely reallocated.

Volatility is likely to remain bid as long as dispersion stays elevated and options markets keep paying up for crash protection. That is usually good for market-makers, structured products desks, and option sellers only if they can warehouse risk cheaply; for everyone else, the bigger edge is in relative value rather than direction. The contrarian point is that some of the move may still be a legitimate repricing of AI earnings power, so the right expression is not blanket shorting of growth, but owning the dispersion and fading the most crowded proxies while keeping exposure to genuine cash-flow compounding.

Key Decisions for Investors

• Buy near-dated CBOE call spreads or long-vol structures into the next 4-8 weeks; elevated dispersion can keep realized vol sticky, and upside is convex if a major AI/IPO event widens breadth again. Risk: theta decay if the tape consolidates.
• Reduce gross exposure to broad beta and rotate into a long-MU / short-XLK style relative-value basket for 1-3 months; semis with direct AI linkage are still the cleaner expression than index-heavy megacap exposure, but size must be capped because leadership can reverse fast.
• Short GS or JPM against a basket of volatility beneficiaries for 1-2 months; underwriting optimism and deal exuberance usually compress underwriting quality before they fully benefit from fee pickup. Risk/reward is best on dips after any strong deal-printing headlines.
• Avoid chasing IPO exposure indiscriminately; if participating, prefer a hedged pair such as long profitable AI infrastructure suppliers vs short pre-profit listing candidates over the first 30-60 days post-IPO. This captures primary-market enthusiasm while limiting valuation gap risk.
• For portfolios with large passive exposure, add downside protection on ACWI or SPY via 1-2 month put spreads funded by trimming crowded winners. The goal is not a crash call, but to insulate against a dispersion unwind that can hit active and passive books simultaneously.