Major U.S. artificial intelligence companies have committed to $3 trillion in future financial obligations that do not appear on their balance sheets. This figure, calculated by The Wall Street Journal, represents commitments for data center leases and microchip purchases, contrasting sharply with the sector's collective market capitalization of $600 billion.
These liabilities are disclosed by companies in their quarterly Securities and Exchange Commission filings but are not yet recorded on balance sheets under Generally Accepted Accounting Principles (GAAP). They become reportable only when certain conditions are met, such as the commencement of a lease or delivery of purchased assets.
Companies such as Alphabet, Microsoft, Meta Platforms and Amazon are among those with significant off-balance-sheet commitments. The $3 trillion includes obligations for AI-capable processing chips, data center networking solutions, electricity to power energy-intensive facilities, and future leases of data centers. Some of the facilities subject to these lease commitments have not yet been constructed.
Companies often opt to lease rather than own data center facilities, preserving flexibility to exit leases—even with penalties—rather than committing to outright ownership costs of massive infrastructure. This year alone, these major companies are budgeting approximately $750 billion for AI infrastructure.
Financial analyst Alex Krueger characterized the situation as "the biggest bubble in stock market history." The initial reporting on these hidden liabilities triggered share declines across implicated companies. Ancillary businesses benefiting from the AI build-out, such as GE Vernova and Vertiv, also saw stock declines in response to the news.
