The Big Short’ investor has placed a $1B wager against what he sees as an AI bubble, claiming Meta and Oracle’s financial reporting masks deeper problems

Michael Burry, known for predicting the 2008 crisis, resurfaced to warn that Big Tech’s AI profits rely heavily on extending depreciation schedules for servers and GPUs—an accounting tactic he says artificially boosts earnings. His comments came just after dissolving Scion Asset Management’s registration, freeing him from reporting obligations and prompting speculation he’s stepping away from what he views as a distorted market.

Burry estimates companies will understate nearly $176 billion in depreciation between 2026 and 2028, inflating profits at firms like Meta and Oracle by over 20%. Analysts such as Dylan Morrow and Richard Jarc support his critique, arguing that AI hardware becomes obsolete far faster than the five- or six-year lifespans companies now claim, especially as Nvidia moves to yearly chip releases.

The Economist recently highlighted that extending depreciation timelines could distort Big Tech’s valuations by as much as $4 trillion. While some, including Bank of America, still see strong AI demand, critics argue that massive AI infrastructure spending is being mistaken for real productivity and that companies are attempting megaprojects they’ve never handled before.

Additional risks come from power constraints: many new data centers sit idle waiting for grid hookups, with billions in equipment depreciating each month. Morrow warns the sector is heading toward overcapacity and weak returns, while investor exposure is dangerously concentrated in a handful of megacap tech stocks—making the market vulnerable to a rapid reversal.