Same AI trade, same problems

Those are valid concerns ahead of a flurry of Big Tech earnings reports due out next week, including Google parent Alphabet, Facebook parent Meta Platforms, Microsoft, and Apple. Investors had gotten antsy about eye-watering AI bills, and they’re eager for any hint that massive spending is slowing—or at least evidence that companies are seeing returns on all that money.

They’re right to be concerned, says a group of Morgan Stanley analysts led by Todd Castagno. While Big Tech companies have deep pockets, they don’t have infinite money, and most of their AI obligations are off-balance-sheet and opaque, since accounting rules often allow them to defer registering liabilities until triggers like delivery or lease commencements.

“The lack of disclosure and contractual complexity of these arrangements makes it difficult for investors to interpret true economic leverage versus that reported on balance sheet,” Castagno’s team writes. “The circularity of the AI ecosystem further complicates adequate analysis.”

The stakes are undoubtedly high. Purchase obligations from hyperscalers and Nvidia have topped $640 billion, more than doubling in the past year and up six times in the past five years. Commitments are up relative to cash flow too, with Meta’s at approximately 1.7 times forward operating cash flow, and Oracle’s more than seven times.

The upshot is that AI commitments are becoming “more frequent, larger, and more complex,” the Morgan Stanley team notes, and regular investors have less ability to “assess companies’ total potential leverage, which is rising much faster than balance sheet leverage.”

Everything works well as long as there are no hiccups. Even so, it’s hard to ignore how intertwined these companies are. Nvidia and hyperscalers like Google promise to rent space in data centers from suppliers, who then secure loans to build those centers from banks and private-credit lenders, who are reassured by Big Tech’s creditworthiness.

None of this is chump change.

“As of the latest disclosures, hyperscalers have $82 billion of finance lease liabilities and $175 billion of operating lease liabilities on their balance sheets, but they have also committed to $675 billion of lease payments to leases that have not started and will remain off-balance-sheet until they begin,” Castagno’s team writes.

Again, tech companies aren’t doing anything wrong or illegal—it’s perfectly fine for many of these obligations to remain off-balance-sheet, but the numbers might rightfully make some investors nervous, especially when these bills aren’t theoretical, but highly likely.

All of this is already an adjustment for tech shareholders, who until a few years ago were invested in some of the world’s biggest and most reliable free-cash-flow generators; today those same companies aren’t exactly penny-pinching, but they are taking on increasing debt.

With AI back in the driver’s seat, these pitfalls may come back into focus too.

Write to Teresa Rivas at teresa.rivas@barrons.com