Big tech companies report huge spending numbers every quarter for artificial intelligence infrastructure. These numbers cover data centers and computer chips.
But the reported figures don't tell the full story. A large part of future spending commitments never shows up on company balance sheets.
A Wall Street Journal analysis found that nine top tech companies have about $3 trillion in off-balance-sheet commitments. Most of this money is tied to AI projects.
These hidden obligations are growing faster than traditional capital spending. Capital expenditures across these companies totaled about $600 billion over the past year.
The off-balance-sheet commitments are about triple what these companies owe under current leases and long-term borrowing combined.
How the Numbers Break Down
Alphabet, Meta, Amazon and Microsoft show some of the largest gaps between what they report and what they actually owe.
Purchase commitments alone add up to about $1.52 trillion across these companies. Leases that haven't started yet total roughly $904 billion.
By comparison, lease liabilities that are already on the books total $248 billion. Long-term debt on the books totals $356 billion.
These are big differences. The off-balance-sheet totals dwarf what shows up in standard financial statements.
Alphabet's total commitments, both on and off balance sheet, reach about $811 billion. Meta's total sits near $349 billion.
The Bet Behind the Spending
These companies are betting that demand for AI computing will keep growing for years to come. They also need enough chips and hardware to meet that demand.
The plan is that future revenue from AI adoption will cover these massive obligations. Companies expect businesses and consumers to use AI tools in daily life.
If demand for AI doesn't grow as expected, these commitments could become a heavy financial burden. That risk would fall on both the companies and their investors.
Meta's Hyperion data center project in Louisiana offers one example of how this works. The facility is about the size of 1,700 football fields.
Projects like Hyperion often get funded through special financial structures. These structures keep the debt off the main company balance sheet, even though the company is still on the hook for the payments.
Nvidia and Oracle also show sharp increases in off-balance-sheet obligations compared with a year earlier. Their growth rates in these commitments outpace their capital expenditure growth by wide margins.
The data comes from footnotes in each company's most recent securities filings. Most companies reported figures as of June, though Nvidia's data is from April and Meta's from July.
Investors are watching these numbers closely as AI spending keeps climbing across the industry. The gap between reported capex and total commitments continues to widen each quarter.
