Analysts say AI cloud giants have large off-balance sheet commitments

In short: Big cloud companies building AI data centers may have large payment commitments that are not shown as debt yet, according to analyst estimates.

What's going on

Big “hyperscalers” like Alphabet, Microsoft, Amazon, and others are spending heavily on data centers, the large buildings full of computers that run AI and online services.

These companies often look financially safe on paper. Morgan Stanley said the group’s average “net leverage” (a rough debt load measure) is about 0.5 times, and many have more cash than debt.

But analysts say a growing share of the money is being raised in ways that do not show up as normal debt right away. One example highlighted by the Financial Times involved Meta and a partner setting up a separate project company to own a Louisiana data center. Meta only owned 20 percent, but agreed to rent the facility for at least 20 years, which helped that project company sell about $27 billion in bonds.

Goldman Sachs analysts estimated the hyperscalers now have about $1.5 trillion in total lease commitments. They said about $1.0 trillion of that is for leases that have been agreed but have not started yet, so they do not appear on the companies’ balance sheets under US accounting rules. Instead, they are often disclosed in footnotes, which some investors may overlook.

Morgan Stanley also counted about $982 billion in “purchase commitments” at the end of the first quarter. These are promises to buy things like chips, equipment, computing capacity, and electricity later.

What to watch

These commitments are like signing a long-term contract before the first bill arrives. If AI demand keeps growing, the spending may pay off. If it does not, investors and lenders may pay more attention to how much future cash these companies will need to meet contracts.

Source: Financial Times