Goldman Sachs Underweight on Hyperscaler AI Debt on Supply Flood

Michael Burry is warning that Big Tech's unprecedented AI infrastructure spending could eventually leave hyperscalers with significant write-offs, arguing that today's buildout is beginning to resemble earlier capital booms that ended with excess capacity and falling returns. Burry said net capital investment by S&P 500 companies reached 2.07% of U.S. GDP as of June 30, the highest level in nearly four decades outside the aftermath of the March 2000 Nasdaq peak, and he wrote that he has little doubt the next few quarters will set still higher marks, possibly even eclipsing that aftermath of the 2000 tech stock peak. He estimates Microsoft, Amazon, Alphabet, Meta Platforms and Oracle have accumulated roughly $3 trillion in purchase commitments, future leases, guarantees, construction-in-progress and other exposures tied to AI infrastructure, and he wrote that when the write-offs come, perhaps in 2028 or 2029, these commitments may be so large that a relatively small write-off has a bigger impact than we can now imagine. Burry pointed to the late-1990s technology, media and telecommunications boom, after which depreciation and write-downs helped push aggregate S&P 500 net investment below zero for 12 consecutive quarters between mid-2003 and mid-2006, and he questioned Oracle's accounting around customer prepayments and future cloud revenue amid scrutiny of the company's debt and Project Jupiter data center. For investors, the critical metrics are free cash flow, capex intensity, utilization, depreciation, lease obligations and returns on invested capital, making 2027 through 2029 especially important for judging whether today's AI buildout becomes a durable productivity cycle or another period of overinvestment.