Commentators compare AI data center debt boom to 2008 risks

In short: More people are warning that today’s rush to build AI data centers is being financed in ways that resemble some of the risky borrowing seen before the 2008 crisis.

What's going on

A growing set of commentaries is using the line “It’s as though 2008 never even happened” to describe worries about how the AI boom is being paid for. The point is not that the 2008 financial crisis is repeating exactly. It is that some funding methods can hide risk until something goes wrong.

The focus is on the physical buildout behind AI, especially large data centers (warehouse-sized buildings filled with computers). Critics point to heavy borrowing to fund construction, and to complex setups that keep debt “off the balance sheet,” meaning it does not show up clearly in a company’s main financial statements (like keeping big bills in a side notebook). Michael Burry has argued that parts of the data center boom involve “circular financing” and other structures that can pile up hidden leverage, which is another way of saying lots of borrowing stacked on top of more borrowing.

Some of the concern is driven by the size of the spending. Bloomberg-compiled data cited by Kitco says companies have borrowed more than $410 billion this year for data centers and other AI-related investments. Senator Elizabeth Warren has also warned that AI investment could help trigger a broader crash, and she has explicitly drawn parallels to 2008.

What to watch

Not everyone agrees the risks add up to a system-wide threat. The Economist has argued fears are overdone, and Nvidia CEO Jensen Huang has rejected doomsday claims. The key question now is whether rising debt and hard-to-see financial arrangements stay manageable as interest rates, demand for AI services, and data center profits change.

Source: NYTimes