As the Stock Market Flashes a Warning Seen Only 2 Times in 155 Years, Warren Buffett has a Big Warning for Investors

The S&P 500 has gained roughly 26%, 25%, and 18% over the past three years, but the article warns that AI-driven enthusiasm has pushed the CAPE ratio to 41, a level previously seen only ahead of the 1929 crash and 1999 dot-com bubble. Goldman Sachs estimates AI accounts for about half of recent S&P 500 earnings growth, while AI infrastructure capex is projected to exceed $1 trillion this year. Warren Buffett cautioned that value is difficult to find when investors are "preferring gambling," underscoring valuation and business-sustainability risks in a concentrated AI-led market.

Analysis

The actionable issue is not an abstract valuation comparison but whether AI infrastructure spending continues to convert into end-demand. NVDA and MU remain levered to hyperscaler capex revisions, yet their risk profiles differ: NVDA is exposed to accelerator unit economics and custom-silicon substitution, while MU is exposed to the more cyclical memory pricing cycle. A deceleration in cloud capex guidance would likely compress both multiples before reported revenue weakens, with MU carrying greater downside beta if HBM supply tightness normalizes.

The consensus bearish extrapolation is also incomplete: elevated index valuation alone has been a poor short-timing signal, particularly when index concentration means a small number of cash-rich platforms fund investment internally. The more relevant 1-3 month catalyst is whether MSFT, AMZN, GOOGL and META disclose rising depreciation, power/network constraints, or weaker AI monetization against sustained capex; that combination would shift the market from "capacity scarcity" to "return-on-capital" analysis. Conversely, enterprise inference revenue or renewed supply constraints would extend the infrastructure cycle and make a valuation-only hedge expensive.

This article is not independently useful as a catalyst and contains promotional framing, so a directional S&P 500 short is not warranted. The better expression is to own dispersion: hedge the premium hardware beneficiaries against firms carrying the capex/depreciation burden, then add risk only after quarterly guidance establishes whether spending remains demand-led rather than competitive signaling. Over 6-18 months, lower-cost custom accelerators and efficiency gains are a structural margin risk for merchant GPU suppliers, but the timing depends on software portability and developer adoption.

Key Decisions for Investors

• Do not initiate a broad SPY short on valuation rhetoric alone; use any index hedge only after a hyperscaler earnings cycle shows simultaneous capex-guide cuts and softer AI revenue commentary. Falsifier: aggregate MSFT/AMZN/GOOGL/META capex guidance remains above consensus while AI monetization disclosures improve.
• Establish a modest 1-3 month relative-value hedge: long MSFT or GOOGL / short an equal-dollar basket of NVDA and MU, sized as a dispersion trade rather than a net market short. The thesis is that platforms retain strategic control and diversified cash flows if hardware ROI is questioned; cover if NVDA backlog commentary or MU HBM pricing guidance accelerates materially.
• For existing NVDA and MU longs, buy post-earnings downside protection rather than reduce solely on macro valuation concerns: 3-6 month put spreads funded with out-of-the-money calls limit carry while protecting a capex-reset scenario. Reassess after the next hyperscaler reporting round; a break in AI infrastructure order visibility, not a CAPE level, is the trigger to increase hedges.
• Monitor MU-specific HBM contract pricing, inventory days, and conventional DRAM/NAND pricing weekly. If HBM mix gains stall while spot memory prices soften, favor a tactical MU underweight versus NVDA; if both HBM supply remains constrained and memory pricing rises, the relative short should be closed.