Michael Burry and Goldman Sachs Asset Management are simultaneously warning that the AI infrastructure investment boom is shifting from free-cash-flow financing to debt and long-term commitments, with credit markets already showing signs of strain. Oracle’s credit default swaps spiked to a record high last Thursday, according to Zerohedge, while Goldman Sachs Asset Management is underweight the largest AI borrowers.
In his latest article, Burry said net capital investment by S&P 500 companies reached 2.07% of GDP as of June 30, the highest in nearly four decades outside the aftermath of the March 2000 Nasdaq peak. He expects net investment/GDP to set higher marks in the coming quarters and possibly even eclipse the aftermath of the 2000 tech-stock peak.
Burry said Microsoft, Amazon, Alphabet, Meta and Oracle have accumulated roughly $3 trillion in purchase commitments, future leases, guarantees, construction-in-progress and other exposures tied to their AI infrastructure buildout. He warned that write-downs could arrive in 2028 or 2029, and that by then the commitments may be so large that a relatively small write-down has a bigger impact than can now be imagined.
His argument centers on the capital cycle: periods of heavy investment can eventually lead to overcapacity, falling returns and write-downs. He pointed to the late-1990s technology, media and telecommunications buildout, saying subsequent depreciation and write-downs were severe enough to push aggregate S&P 500 net investment into negative territory for 12 consecutive quarters between mid-2003 and mid-2006. Burry believes a similar dynamic could emerge from the current AI infrastructure boom. In his view, the issue is not simply the amount of capital being deployed, but the scale of future commitments being accumulated by the largest hyperscalers.
“The most profitable companies in the world, except Apple, are betting everything on this as their free cash flow turns negative and the borrowing pace increases,” he wrote.
Burry also questioned Oracle’s accounting around customer prepayments, highlighting a nearly 20% increase in future cloud revenue without delivering extra service or work product immediately, but “created by the structure of the contract.” Credit markets are also flashing caution. Oracle’s CDS spiked to record highs last Thursday, signaling mounting investor anxiety over the AI hyperscaler’s debt, according to Zerohedge. Bloomberg reported the company is moving to shield itself from rising expenses on its massive Project Jupiter data center in New Mexico, sending a force majeure notice to the project’s developer.
At Goldman Sachs Asset Management, Lindsay Rosner, head of multi-sector fixed income investing, said the firm is underweight the biggest AI borrowers as a flood of issuance washes over the market. “We believe there will be a lot of hyperscaler issuance,” Rosner said in a Thursday interview on Bloomberg TV. “For that sector at large we are underweight knowing more issuance will come.”
Big Tech companies have been among the biggest sellers of high-grade corporate bonds in the U.S. this year, with Amazon, Meta and Alphabet borrowing tens of billions of dollars for AI projects. “That is what we have seen all year long,” Rosner said. “New issuance is replacing existing issuance.” Rosner said the underweight call reflects expectations for heavy supply, rather than pessimism about AI. Her team believes “in the story of AI,” she said, but is “cognizant” that more borrowing could reprice the market.
Major hyperscalers—including Alphabet, Meta, Amazon, Microsoft and Oracle—have increasingly relied on debt to fund an unprecedented capital expenditure surge as their free cash flow turned negative. According to new Brookings Institution research, the AI infrastructure boom could require $10.3 trillion of U.S. investment through 2032, with a growing share of the financial risk shifting away from Big Tech balance sheets and into less-transparent financing structures.