In a recent social media post, Burry pointed to a subtle but powerful move by companies like Meta and Alphabet. By lengthening the depreciation schedules for servers, chips, and other AI infrastructure, these companies are lowering the non-cash expenses that weigh on net income. That means higher profits on paper, without any real increase in cash flow.
“These accounting changes are not illegal, but they do mask the actual cost of operations,” Burry warned. He recently terminated his fund’s registration with the Securities and Exchange Commission, adding further intrigue to his critique.
Meta now estimates its computing gear lasts five and a half years, up from four. That change alone reduced its 2025 depreciation expense by nearly three billion dollars. Microsoft and Alphabet have followed suit, claiming extended usefulness of their hardware. The impact is significant. Depreciation expenses fell, profits rose, and investors cheered.
But not everyone is buying it. Amazon took the opposite route, shortening its server life from six to five years, citing the rapid pace of chip upgrades.
“AI hype is shifting into a phase where it has to prove itself financially,” said Anthony Saglimbene, chief market strategist at Ameriprise Financial. “Investors are asking harder questions.”
Capital expenditures are exploding. Meta, Alphabet, Amazon, and Microsoft are expected to spend more than four hundred sixty billion dollars on AI infrastructure over the next twelve months. Most of that will go toward equipment that loses value quickly.
Even with the new accounting, depreciation costs are climbing. Alphabet, Meta, and Microsoft recorded twenty-two billion dollars in depreciation last quarter, up from ten billion a year ago. That figure is projected to hit thirty billion by this time next year.
Yet third-quarter earnings still beat expectations. The Magnificent Seven are on pace for twenty-seven percent profit growth, nearly double earlier forecasts.
Some analysts remain bullish. Others, like University of North Carolina professor Stephen Glaeser, say the risk is real. “If AI returns do not accelerate, these assets become dead weight,” he said.