Credit markets price AI blowup at 2008 levels while Burry rolls Nvidia shorts and Mag7 returns keep shrinking

by
Share this story

Michael Burry sold his Microsoft longs and closed Oracle shorts this week.
He rolled Nvidia puts out to June 2027 and QQQ puts to February 2027.
He still holds shorts on Nvidia SOXX Micron Caterpillar Palantir Tesla and Applied Materials.
Oracle 5-year CDS hit 215 basis points higher than any level in 2008.
Hyperscalers face over $1 trillion in uncommenced data-center leases and commitments not yet on balance sheets.
Microsoft alone listed $329 billion in leases that had not commenced.
Magnificent 7 total annual returns have gotten smaller every year since 2023.
AMD reported record $11.5 billion revenue up 50 percent with data-center sales doubled yet the stock dropped 7-9 percent.





AMD’s revenue climbs 50% and data center sales doubled, but the stock is down

Revenue Is Not Profit, and Investor Money Is Not Sustainable Revenue. The shaky ground under AI capital flows
byu/zolo_black inAIBubble

It feels to me that the market is blinded by the huge revenue growth at the hyperscale clouds and treat it as proof that AI is a success story and there is strong demand for AI and that the capex is going to be profitable.
I think they are missing that a lot of that revenue is coming from unprofitable AI labs and unprofitable, moatless AI startups spending money they raised from investors, lenders, capex, and circular financing.

Unfortunately, I don’t have the numbers to know how much from AI cloud revenue comes from investors and lenders and how much comes from end users.

I think the biggest question is which of these two happens first. Either the money flow from one of the sources in the diagram dries up for some reason (an interest rate spike, a bear market, private equity firms going bankrupt, and so on), or real profits, not just revenue, start coming from real users and make all of these loans and investments worth it.

I see endless possibilities for something to go wrong in this capital flow, and I don’t see how it can generate real profit before it collapses.

Note: this is only about the AI part. I’m not talking about the normal cloud business or the rest of the hyperscalers’ profits that have nothing to do with AI. Those are real and profitable. This is only an attempt to look at the AI piece.

A lot of pieces are missing from the capital diagram. I tried to focus on the things I thought were most relevant.

 

0 views