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

The article warns that the S&P 500's roughly 26%, 25%, and 18% gains over the past three years have been heavily driven by AI, which Goldman Sachs estimates contributed about half of recent index earnings growth. AI capital spending is expected to exceed $1 trillion this year, but OpenAI reportedly lost $3.7 billion in Q1 2026 on $5.7 billion of revenue, underscoring uncertainty around downstream business-model viability. The CAPE ratio stands at 41, a level previously seen only ahead of the 1929 crash and the 1999 dot-com peak, while Warren Buffett cautioned that values are difficult to find amid speculative behavior.

Analysis

The relevant risk is not broad AI adoption but a narrowing return-on-invested-capital spread across the stack. NVDA and MU remain leveraged to hyperscaler capex, yet their equity sensitivity rises sharply if customers shift from capacity acquisition to utilization discipline: a modest reduction in 2027 accelerator orders would flow through inventories and gross-margin expectations faster than end-demand. The more vulnerable second-order cohort is AI-adjacent infrastructure priced on sustained double-digit capex growth—networking, power/cooling and smaller semiconductor names—where backlog quality and customer concentration are weaker than at the leaders.

Near term (days to weeks), valuation warnings alone are unlikely to break the AI complex while earnings revisions and liquidity remain supportive; this is a positioning issue, not a discrete fundamental catalyst. Over 1-3 months, the key catalyst is hyperscaler guidance on incremental AI revenue versus depreciation and capex, especially any indication that inference monetization is lagging compute deployment. Over 6-18 months, a capex digestion cycle would favor cash-generative software and platforms that can monetize AI without matching infrastructure spend, while compressing semiconductor multiples even if absolute revenues continue growing.

The contrarian view is that a high aggregate market valuation is a poor timing signal: concentration can persist if earnings breadth broadens outside AI. The more actionable asymmetry is dispersion within AI—own the firms with pricing power and balance-sheet capacity, hedge suppliers whose earnings require uninterrupted capex acceleration. Buffett-style caution should not be expressed as an unhedged index short absent deterioration in earnings revisions, credit spreads, or labor-demand data.

Key Decisions for Investors

• Maintain NVDA exposure only against a defined semiconductor hedge: long NVDA / short SOXX in equal beta through the next hyperscaler earnings cycle. NVDA’s software ecosystem and customer priority allocation should outperform in a capex slowdown; exit if NVDA’s data-center revenue guidance decelerates materially while SOXX relative strength holds.
• Do not add MU on AI narrative alone before the next earnings release. Initiate only if HBM pricing, supply discipline, and fiscal-year gross-margin guidance support earnings estimates; otherwise MU is a high-beta short candidate versus NVDA because memory pricing reverses more abruptly in a digestion cycle.
• Buy 3-6 month QQQ put spreads rather than outright index shorts if portfolio AI beta is elevated. Structure strikes around a 7-12% downside range to fund protection against multiple compression while avoiding open-ended carry; reassess after the next CPI, payrolls, and mega-cap guidance cycle.
• Watch hyperscaler capex guidance, NVDA lead times, and semiconductor inventory days as falsification data. A second consecutive quarter of rising capex alongside improving AI-related revenue disclosures invalidates the near-term digestion thesis and argues for covering relative-value hedges.