The race to build AI infrastructure has driven record spending on GPUs, servers and data centers. But increasingly, analysts are looking beyond quarterly capital expenditures to something less obvious: contractual commitments that lock companies into future spending—often hidden in the footnotes of financial filings.

Morgan Stanley estimates that Nvidia Corp. (NVDA) and five hyperscalers—Alphabet Inc. (GOOGL), Amazon.com Inc. (AMZN), Meta Platforms Inc. (META), Microsoft Corp. (MSFT), and Oracle Corp. (ORCL)—have accumulated roughly $1 trillion in purchase commitments that remain off the balance sheet until the related goods or services are delivered under current accounting rules.

Moody's reached a similar conclusion. Earlier this year, the ratings agency estimated approximately $969 billion in future hyperscaler lease commitments, including roughly $662 billion tied to leases that had not yet commenced and therefore had not yet become recognized liabilities.

Meanwhile, the Bank for International Settlements (BIS) has described many of the financing structures supporting AI infrastructure as "shadow borrowing," arguing that they create obligations that are economically similar to debt even if traditional leverage metrics do not fully capture them.

Why Investors Should Read The Footnotes

The discussion gained fresh attention after Nikkei analyzed filings from Alphabet, Amazon, Meta, Microsoft and Oracle, estimating roughly $1.65 trillion of AI-related commitments that remain off the balance sheet under current accounting rules. While that estimate comes from Nikkei's own analysis and was not independently verified by the companies, it aligns with the broader trend identified by Morgan Stanley, Moody's and BIS.

Companies disclose these commitments in regulatory filings under purchase commitments, lease obligations and other contractual commitments. And they typically receive far less attention than reported debt, earnings per share or quarterly AI capex.

That may be worth remembering this earnings season.

Alphabet, Microsoft and Meta will likely reaffirm aggressive AI investment plans. Investors will undoubtedly scrutinize how much they spent last quarter. But the filings may reveal something equally important: how much AI spending has already been committed for the years ahead—and that story is more likely to be found in the footnotes than in the balance sheet.