US tech giants have $3.2tn in hidden AI obligations, and someone has to make them pay back

Nikkei Asia reported this week that US tech giants' off-balance-sheet debt from AI-related obligations has almost doubled in six months, to about $3.2tn. Kohei Yamada, reporting from Palo Alto, puts the rise at 90% and says the commitments and obligations of Google, Meta and other large technology companies now reach 80% of their assets. Nikkei also reports that investors are increasingly alarmed by spending that shows no sign of ending.

The bill sits off the balance sheet

The key word in Nikkei's headline is "hidden". These obligations do not appear on the balance sheet, which is the first place most readers look to judge how stretched a company is. The photograph Nikkei ran with the story is of an Amazon data centre in Indiana. That is where the money goes: infrastructure the tech giants have promised before they have paid for it.

The speed matters more than the size. An earlier Nikkei headline put the hidden debts of five US tech giants at $1.65tn, attributed to opaque AI funding. This week's figure is close to twice that. The same Nikkei page lists two more signs of how far the funding effort now reaches: Anthropic seeking up to $3.5bn from Japanese investors in its IPO, and yen bond offerings hitting a record $16bn as the AI funding race reaches Japan.

When obligations equal 80% of assets, Google, Meta and their peers have already made the decision. They have committed to AI at a scale they cannot quietly reverse.

What a committed balance sheet means for the people inside it

The question we draw from Nikkei's figures is where the return comes from.

A company that has signed obligations on this scale needs them to pay back. There are two ways to do that. AI products bring in new revenue, or AI lowers the cost of running the business. The second way goes through people's work: tasks are absorbed, teams get smaller, and roles thin out one restructuring cycle at a time.

This pressure does not stay inside Google or Meta. Every company that buys their cloud capacity, licenses their models or sells into their supply chain gets pulled into the same logic. A vendor that has committed this much capital needs its customers to adopt quickly and widely. Customers then need adoption to appear in their own numbers.

For an operator, this changes what the next performance review measures. Leaders who justify AI spending will look for evidence that it is working. Usage counts are the easiest evidence to collect. Improved results are the most useful. The person who can show improved results with numbers is the one the spending case depends on.

AI does the routine work. You do the thinking. The companies carrying $3.2tn in obligations need that thinking to produce returns, and they will reward the people who can show it.

Your Next Move

1. Read your employer's exposure. If your company is listed, find the commitments and obligations note in its latest annual report and read it alongside the debt line. If you work for a supplier or customer of the large technology firms, estimate how much of your revenue depends on their AI spending continuing at its current pace. Plan for the case where it slows.

2. Put a number on one AI workflow before your next review. Choose one recurring piece of your work. Run it with AI for a month. Record hours saved, errors reduced or output gained. Our guide to showing measurable impact when your employer grades AI use explains how to present that evidence and what the law gives you.

3. Watch the funding signals each month. Track Nikkei's reporting on these obligations and on new fundraising such as IPOs and bond issues. If the obligations keep rising faster than revenue, cost pressure will reach headcount decisions. Make your measured results visible before that happens.