• Veteran strategist David Rosenberg warns that U.S. stock valuations, leverage, and investor positioning have reached unprecedented extremes in his four-decade career, and he is shifting toward U.S. Treasuries and international equities.
  • Rosenberg argues the market is more dependent on AI than stocks were on technology before the dot-com bubble, with narrow leadership masking weakness across the average S&P 500 constituent.
  • The warning is a strategist's assessment, not an established finding that a crash is imminent, and it comes as higher bond yields and a demanding earnings season test the AI trade.

A Historic Extreme

David Rosenberg, founder of Rosenberg Research and former chief North American economist at Merrill Lynch, has issued one of his starkest warnings yet: U.S. stock valuations, leverage, and investor positioning have reached extremes he has not witnessed in his four-decade career. In comments reported by Bloomberg on October 13, Rosenberg argued that the market's dependence on artificial intelligence now exceeds the pre-dot-com reliance on technology stocks. "I'm looking at stuff I haven't seen before," he said, according to the report.

The warning is as much about concentration as it is about expensive stocks. Rosenberg contends that U.S. markets, investor portfolios, and even consumer confidence have become unusually dependent on the AI boom, a dynamic he says leaves the broader economy vulnerable to a reversal in a handful of mega-cap names.

Breadth Tells a Different Story

Rosenberg has been building this argument for weeks. On October 6, he noted that Nvidia (NVDA)'s strength was masking weakness elsewhere: the average S&P 500 constituent was approximately 20% below its 52-week high, while the median constituent sat about 17% below. The divergence is striking. September's capitalization-weighted index gained just 0.2%, while its equal-weighted counterpart fell 4.4%, according to reporting citing MarketWatch. That gap supports his concern that headline indexes overstate the health of the typical stock.

The AI trade is also facing a more demanding earnings season. Barron's reported on October 13 that expensive stocks—including AMD (AMD), Marvell (MRVL), Corning (GLW), and Bloom Energy (BE)—suffered unusually large valuation contractions during the third quarter. Investors increasingly want evidence that future AI revenue will justify today's massive infrastructure spending.

A Qualification Worth Noting

There is an important caveat to the headline. The broad market's valuation has already moderated. Barron's, citing Dow Jones Market Data, reported that the S&P 500's forward price-to-earnings ratio fell to about 19 from 22 at the start of 2026. Rosenberg's "most stretched" characterization, therefore, should not be read as proof that every conventional valuation measure sits at a 40-year maximum.

Higher bond yields are central to the debate. The U.S. 10-year Treasury yield finished the preceding week above 5%, making government bonds more competitive with equities. Higher financing costs also raise the hurdle for debt-funded AI projects and reduce what investors are willing to pay for profits expected far in the future. Rosenberg's reported shift toward Treasuries and international equities is, in part, a diversification argument—not evidence that overseas markets are insulated from U.S. technology demand or geopolitical risk.

The Other Side of the Trade

The bearish case is hardly unanimous. October 12 Bloomberg reporting described one of the strongest corporate profit cycles in recent history. Goldman Sachs (GS) strategists raised their three-month equity view to overweight, while E*Trade's Chris Larkin emphasized AI's longer-term tailwinds. These assessments directly challenge the idea that elevated risk necessarily means an imminent downturn.

Regulatory developments add another layer. Reuters reported on October 13 that banking regulators, including the Federal Reserve, were questioning banks about advanced AI, Chinese-developed models, autonomous agents, and cybersecurity controls. Separately, an October 1 New York Fed analysis found that U.S. suppliers experienced an average cumulative abnormal return of roughly −3.6% after export controls targeted their Chinese customers. National-security policy can therefore constrain commercial opportunities for firms participating in the AI supply chain.

What to Watch

In the short term, earnings releases, AI spending guidance, inflation readings, and Treasury yields are the principal tests. Strong results could support valuations; disappointing monetization could accelerate a rotation away from expensive AI-linked shares. The economic exposure extends beyond stock prices. Federal Reserve research describes the AI infrastructure boom as a significant driver of global investment and U.S. imports of high-technology capital goods. A slowdown could therefore affect equipment suppliers and international trade, not just American technology shareholders.

Rosenberg's warning matters most to retirement savers and index investors because weak performance across many stocks can coexist with resilient headline indexes when a few large companies dominate returns. His consumer warning invokes the wealth effect: households relying on rising portfolio values could become more cautious if markets fall. That remains a risk scenario, not a verified forecast of declining consumption.

The evidence supports taking concentration and financing risks seriously. It does not establish Rosenberg's superlative as a universally accepted statistical conclusion, nor does it provide a reliable timetable for a market reversal. Attempts to reach Rosenberg Research for further comment were not immediately successful.

Clarification: An earlier version of this article did not specify that the 20% and 17% figures refer to the average and median S&P 500 constituent, respectively, rather than the index itself.