Warren Buffett Warned Investors About This Number 27 Years Ago. It's Never Been Higher. Should AI Investors Worry?

After-tax corporate profits reached about 13% of U.S. GDP in Q2 2026, a record in government data dating to 1947; four major technology companies spent roughly $202 billion on property and equipment in the quarter ended Sept. 30, more than double a year earlier. The article cautions that much AI revenue depends on spending within the industry, while higher labor costs and interest rates could pressure profits; CoreWeave bonds were yielding more than 16%, and the Nasdaq 100 fell 6.5% Thursday before gaining 1.4% Friday. Berkshire Hathaway reported a record quarterly $9.6 billion stock buyback, bringing its year-to-date total to about $14.4 billion.

Analysis

The key risk is not that AI demand is imaginary, but that current supplier earnings are unusually exposed to a small group of customers whose capex is discretionary. That creates reflexivity: a pause in hyperscaler spending can hit accelerator orders and rented-compute utilization before the spenders’ own earnings benefit from lower capex. Conversely, if AI services generate meaningful external revenue, today’s circularity concern will prove too bearish.

The exposure is asymmetric. Amazon, Alphabet, Microsoft and Meta can defer projects and preserve cash; specialist infrastructure providers have less flexibility, and leverage makes utilization or refinancing disappointments more consequential. Nvidia is exposed to customer budgets, but its position is not equivalent to a leveraged compute lessor. Berkshire’s repurchases are a valuation signal, not evidence that the whole market is expensive or that its intrinsic value is assured.

Near term, treat the recent Nasdaq drawdown as a volatility warning, not confirmation of a profit-share reversal. Over 1–3 months, hyperscaler capex guidance and Nvidia’s next report are the key tests. Over 6–18 months, rising wage costs or tighter financial conditions could pressure margins more broadly; that thesis needs evidence in labor costs and guidance, not just a record aggregate profit share. The contrarian risk is underestimating durable AI monetization and treating all capex as waste. No blanket tech short is warranted.

Key Decisions for Investors

• Prefer a modest relative-value short in CoreWeave versus a long in a diversified, cash-generative platform such as Microsoft or Alphabet; use defined-risk structures if available. Enter only if CoreWeave’s elevated bond yields persist or worsen and its cash conversion fails to validate contracted demand. Exit/reassess if refinancing conditions ease and utilization or external customer revenue improves.
• Do not short Nvidia solely on the circular-spending argument ahead of its next report. Use the November 17 report as a catalyst check: cut exposure or hedge if customer capex plans soften, and reconsider bearish positioning if demand broadens beyond hyperscaler budgets.
• Keep any broad AI-beta reduction gradual; a rotation toward Berkshire Hathaway (BRK.A) is a lower-beta alternative, not a direct hedge. Reassess if hyperscaler guidance shows capex discipline without slowing monetization, or if labor-cost and rate pressure fail to appear in margins and guidance.