Sky-high returns and earnings expectations could mean the end is near for this stock-market bull run

Strategists warn that S&P 500 price and earnings have been exceptionally stretched versus historical trends, even as Wall Street expects semiconductor/AI-bottleneck earnings growth to persist for “at least the next few years.” Risks are rising for hyperscalers’ data-center buildout as Chinese AI model progress accelerates and local U.S. opposition to new construction increases. Net: AI-linked upside expectations may be vulnerable, suggesting a more cautious positioning backdrop for the bull run.

Analysis

When both valuation and earnings expectations are extended, the trade is no longer about whether AI is real; it is about whether the market is paying upfront for several years of uninterrupted capex growth. That sets up an asymmetric regime for the most crowded beneficiaries: semis, networking, and hyperscaler suppliers can keep printing decent quarters and still underperform if management commentary implies even a small slowdown in 2025 spend, because duration-sensitive multiples can compress before earnings roll over.

The second-order risk is that AI buildout is increasingly constrained by non-financial bottlenecks: permits, local opposition, grid interconnects, and power availability. If those frictions bite, the near-term winners may actually be the large platforms with the best negotiating leverage and the strongest free-cash-flow optics, while the losers are the equipment names and smaller infrastructure vendors that need continuous order acceleration to justify premium multiples. The market may be underestimating how quickly "good news" on earnings becomes a liability once expectations get too far ahead of observable deployment capacity.

Catalyst-wise, the next 1-3 months matter more than the next 2-3 years: one softer capex guide, one delayed data-center timeline, or one signal that Chinese model efficiency is reducing incremental compute demand could reset positioning. Contrarian view: the current setup may be less about a coming earnings recession and more about an earnings air pocket in the highest-beta AI names, where even unchanged fundamentals can produce material downside if the market stops paying for perfection.

Key Decisions for Investors

• Hedge crowded AI exposure tactically: buy 1-3 month put spreads on SMH or SOXX into strength, targeting a 2:1 to 3:1 payoff if capex guidance cools or multiples de-rate.
• Reduce outright beta in QQQ/XLK and rotate to a barbell of higher-quality cash-return names; the first leg down is likely multiple compression, not an immediate earnings collapse.
• Pair trade idea: short a basket of the most capex-sensitive AI infrastructure names against long software/platform names with more recurring revenue and less dependency on incremental physical buildout, as a way to isolate the expectation reset.
• Set an alert for hyperscaler capex commentary over the next earnings cycle; if forward spend guidance is flat-to-down, expect the AI basket to trade on sentiment rather than fundamentals for several weeks.
• If no new data confirms a slowdown, do not force a structural short here; treat this as a hedge-then-watch setup rather than a high-conviction bearish outright.