Knock-On Effects Could Make This Worse Than The Dot-Com Bust | David Rosenberg

David Rosenberg argues that generative AI is transformative, but investor behavior and valuations have created a bubble: AI investment would need to generate revenue growth of 50% annually for five years to produce a decent return on the hundreds of billions of dollars being committed. He also says hyperscalers are relying on leverage and debt markets rather than corporate cash flow, pushing real rates higher and pressuring credit-sensitive sectors such as housing.

Rosenberg describes economic growth as fragile, slowing from roughly 3% toward 2% with trailing real GDP growth potentially near 1.5%. He attributes the apparent resilience largely to AI data-center construction and equity-wealth-funded consumer spending, while nontechnology business investment is flat and technology spending is up 16%.

Rosenberg argues that roughly $2 trillion in post-COVID stimulus checks delayed the recession many expected but has now faded, making the belief that the business cycle was repealed particularly vulnerable. He also cautions that private final sales near 3% may overstate economic strength because the measure excludes foreign trade even though the AI boom relies heavily on imported inputs.

Rosenberg attributes roughly 60 basis points, or about three-quarters, of the rise in 10-year Treasury yields to the Federal Reserve's shift from expected rate cuts toward possible hikes. He says that communication pushed the market-implied odds of a mid-September hike from about 30% to 60%, while unclear guidance on achieving the 2% inflation target discourages investors from buying duration.

Rosenberg argues that AI-related spending is narrow rather than broad-based: technology investment is up sharply while the rest of corporate spending is flat year over year. He also says private final sales can look stronger than underlying activity because AI's reliance on imported inputs is not fully captured in that measure.

The AI trade has created significant equity concentration, with 40% of the S&P 500 concentrated in 10 stocks, eight of them tied to the same trade, while only healthcare and consumer staples are uncorrelated with it. Rosenberg points to Caterpillar as an indicator of the data-center construction cycle and notes that its stock had fallen more than 25% over the preceding few months.

Rosenberg says U.S. equities are stretched, citing a CAPE multiple near 40 and a negative equity risk premium, with a 2.5% earnings yield versus a 3% real long-bond yield. The speakers nevertheless note that late-cycle conditions do not require an immediate reversal, even though breadth has recently weakened after earlier improvement in small caps, financials, and healthcare.