The Market Is Mispricing A Correlation Shock | Dean Curnutt

Curnutt argues that unusually cheap 6%–7% funding for hyperscalers such as Meta is encouraging AI capital expenditure even as mortgage rates remain near 7% for households. He highlights Anthropic’s potential $2 trillion valuation alongside its CEO’s call to slow the sector as a notable risk signal, not a settled prediction.

Curnutt says geopolitical developments are particularly consequential because crude oil feeds directly into gasoline, plastics, and fertilizer costs while also influencing monetary-policy decisions. He argues that monetary policy cannot credibly control the resulting inflation without more credible U.S. fiscal policy, although he acknowledges limited political space for deficit reduction.

The episode uses 2007–08 to illustrate how risks can emerge in credit before equities: credit markets deteriorated by mid-2007 while stocks continued rising to an October all-time high, and banks’ legacy positions later impaired their ability to service clients. Curnutt presents this as evidence that asset classes and institutions can respond to the same risks at different speeds.

Low realized stock correlation—generally around 5%–15% and sometimes near zero—has reduced index-level volatility and the VIX while creating diversification for index holders. Curnutt contrasts that environment with crisis correlation of roughly 75%–90%, when individual stocks move in tandem with the S&P 500.

The gap between roughly 45% one-month realized volatility for the S&P 500’s top 10 stocks and 7.5% for the index reflects diversification effects, including Microsoft’s negative correlation with six other Magnificent 7 members. Curnutt warns that this apparent diversification could disappear in an AI-led selloff because the top seven or eight stocks comprise more than 35% of the index.