Jamie Dimon sees ‘exuberance’ in markets. That’s a loaded word when it comes to bubbles popping

Jamie Dimon warned that markets may be showing "too much exuberance," especially around AI and Big Tech valuations, echoing Greenspan-style bubble concerns. Panmure Liberum said the AI boom is already about 60% larger than the late-1990s TMT bubble on a macro basis and may require $2 trillion to $5 trillion in annual revenue to justify hyperscaler capex. The piece argues that AI optimism is colliding with worsening macro headwinds and weaker traditional hedges, raising risk of a sharp repricing if growth disappoints.

Analysis

The important read-through is not simply that AI multiples are rich; it is that the marginal buyer is increasingly tied to capex narratives that are only self-justifying if hyperscaler monetization arrives much faster than the market currently assumes. That creates a reflexive setup: suppliers of chips, networking, power, and data-center infrastructure can keep winning on orders even as the eventual equity returns of the platform layer deteriorate. In other words, the near-term earnings line can remain strong while the forward multiple silently compresses, which is the classic late-cycle danger in capex-led booms.

The bigger macro risk is that this is now a concentration trade on U.S. growth. If technology investment is carrying most of real GDP growth, any slowdown in AI spend would not just hit semis and software sentiment—it could bleed into labor demand, industrial utilization, freight, utilities, and bank loan growth with a lag of 2-4 quarters. That means the downside path is less a clean “tech correction” and more a broad de-risking episode where cyclical and defensive correlations both fail, especially if rates stay restrictive and duration stops cushioning equity multiples.

The contrarian point is that a bubble warning can be correct too early, and the hottest names may continue to work until financing conditions tighten or one of the marquee model/equipment vendors disappoints on monetization. The market may be underestimating how long hyper-scalers can sustain elevated capex because they are funding it from operating cash flow rather than external leverage. But that also means the eventual unwind could be slower and nastier: not a funding crisis, but a capital allocation reversal that hits supplier order books first and EPS revisions later.

Key Decisions for Investors

• Reduce gross exposure to AI infrastructure beta: trim NVDA/SMCI/ANET strength into rallies over the next 1-3 weeks; keep any residual exposure as a paired long against software names with clearer near-term monetization
• Pair trade: long JPM vs short DB over 3-6 months. JPM should benefit if volatility rises and capital-markets/liquidity demand increases, while DB is more exposed to a Europe growth scare if AI spend rolls over.
• Buy 3-6 month put spreads on QQQ or XLK financed by selling lower strikes; structure for a 1.5x-3x payoff if AI multiples de-rate 10-15% without needing a full market correction.
• Long utilities/power grid beneficiaries selectively only if AI capex remains intact; otherwise fade the trade. Use tight stops because the second-order demand link to data centers breaks quickly if hyperscaler budgets slow.
• Watch for the first hyperscaler capex guide-down as the catalyst. If one of AMZN/MSFT/GOOGL/Meta signals capex normalization, use it to add to short basket exposure in semis, data-center REITs, and equipment names over the following 2-8 weeks.