Microsoft just delivered a big blow to Michael Burry’s AI bear case

Microsoft’s chief financial officer, Amy Hood, just offered some intel that severely undercuts Michael Burry’s argument against AI stocks, albeit with one big caveat.

If you’ll recall, the hedge fund manager turned Substacker of “The Big Short” fame said that tech companies were understating depreciation charges — that is, how fast GPUs lose their value over time, in a bid to artificially juice profits.

During Microsoft’s conference call on Wednesday, the CFO was asked how the company will be able to capture enough revenue over the six-year useful life of the hardware to justify the outlays. Her response:

“The way to think about that is the majority of the capital that we’re spending today and a lot of the GPUs that we’re buying are already contracted for most of their useful life,” she said. “And so a way to think about that is much of that risk that I think you’re pointing to isn’t there because they’re already sold for the entirety of their useful life.”

The implication here is that not only will these chips make money for as long as tech companies expect they will, but that their useful economic life might actually be longer than that, not shorter.

This tidbit is obviously positive for the hyperscalers, which are spending hundreds of billions on these GPUs. But it’s probably even more of a relief to neoclouds that are even more dependent on these chips being able to generate cash. That’s (mostly) all there is to their businesses, unlike megacap tech giants.

It also corroborates commentary from one such neocloud, CoreWeave, on how well these processors retain value.

“For example, in Q3, we saw our first 10,000-plus H100 contract approaching expiration,” CoreWeave CEO Michael Intrator said after the firm’s most recent earnings report. “Two quarters in advance, the customer proactively recontracted for the infrastructure at a price within 5% of the original agreement.”

And per Silicon Data, H100 rental rates have firmed significantly since the end of November.

However, I’d be remiss not to point out a potential fly in the ointment here: one reason that Microsoft’s GPUs are contracted for most of their useful life is thanks to demand from OpenAI, which accounts for 45% of its commercial remaining performance obligations.

And, if Oracle’s shown us anything, it’s that customer concentration and quality matters.