AMD Just Passed the AI Demand Test. The Financing Test Is Still Open.

A good answer to half the question

AMD had a simple job on Tuesday: show that AI spending is still turning into actual chip sales. It did. Data center revenue nearly doubled, and the stock jumped 8.3% after hours.

That settles the narrow question of whether customers are still buying. It doesn't settle whether they can keep paying, or what happens once chips stop being scarce. If you own this trade, those are the questions that decide how it ends.

"The debate shifted from capability to financing, and it took real casualties among the companies carrying debt to fund the buildout." — Bryan Hayes, Zacks Investment Research

What AMD reported

Data center revenue climbed 97% from a year earlier to a record $8.5 billion, clearing the $8.3 billion that Zacks strategist Bryan Hayes said would suggest AI spending "is landing on real silicon and real revenue." Total revenue of $13.4 billion and gross profit of $7.2 billion both beat expectations. AMD has now topped analysts' revenue and profit estimates four quarters running.

The forecast is the bigger tell. AMD expects $17 billion in fourth-quarter revenue, against an average estimate of $16.2 billion. By my math, that's about 27% above the third quarter, a big step for three months, and it means management sees the strength continuing.

Don't get carried away by the pop, though. AMD shares had fallen 23% in the prior month, so it would take about a 30% gain, not 8.3%, to get back to where they started. Analysts' average price target of $655 implies 40% upside over the next year, but that's a measure of Wall Street's mood, not proof the stock will get there.

Does this rebut Intel? Mostly, and for a telling reason

Intel's stock lost nearly 29% in a day after it forecast $15 billion to $16 billion in fourth-quarter revenue. But look at what it blamed: a shrinking PC market, with rising memory costs squeezing manufacturers and global shipments down 21% in the third quarter, per Counterpoint Research. Its data center and AI segment grew 78% and made $3 billion in operating profit. So Intel's forecast was never strong evidence of an AI slowdown, and AMD's numbers back that up. AMD just has less PC baggage, with nearly two-thirds of its revenue coming from data centers.

The gains also aren't limited to one company. AMD and Intel are both growing fast in data centers, and TSMC reported record third-quarter revenue of $46.7 billion, up 50% from a year earlier. CPUs, the general-purpose processors AMD and Intel make, have gained value because agentic AI, software that carries out tasks on its own, needs that kind of steady computing. This doesn't look like a market where only Nvidia eats.

There's a catch. Intel's CEO says his company can supply only about half of what customers want. When chips are that scarce, nearly every credible seller looks like a winner, so today's strength may partly reflect a shortage rather than lasting demand.

Two things would help show whether this lasts, and neither is in the coverage. One is customers: without a breakdown of AMD's biggest buyers, you can't tell whether a handful are carrying this growth, which would make sales more vulnerable if one pulled back. The other is margins: $7.2 billion of gross profit on $13.4 billion of revenue works out to about 54%, meaning AMD keeps roughly 54 cents of each sales dollar after making its products. Without a prior-period figure or a margin forecast, that's a starting point, not a trend. If you own the stock, check AMD's quarterly filing and call transcript for both.

The bill, and the day supply catches up

Chip revenue is someone else's spending, and the AI industry is borrowing heavily to fund its buildout. It has raised $300 billion in high-grade debt so far this year, per Bloomberg, roughly triple last year's amount. Hayes put it plainly: the debate has moved "from capability to financing."

Oracle shows the strain. It's raising $10 billion in bonds and $10 billion in stock, and says that's the last new funding it needs through 2030. It carries $125.34 billion in borrowings against $67 billion in fiscal 2026 revenue, and its BBB- rating sits one notch above junk. The stock fell 9% on Monday. Orders for the bonds peaked at $38 billion, so lenders haven't left. They're just charging more, and Oracle doesn't expect positive free cash flow (the cash left after bills and investment) until the six months ending May 2028.

The worry goes beyond Oracle. S&P put Meta's AA- rating on negative outlook, saying its ability to monetize AI is "more limited than that of its peers," though a one-notch cut would still leave Meta comfortably high-grade. That's still awkward for the chip story, since Meta's Muse is a leading example of the agentic AI lifting CPU demand. CoreWeave's credit default swaps, which work like insurance against a default, recently priced in a 60% chance of one, according to ICE Data Services. The coverage doesn't show how much of AMD's revenue traces back to stretched borrowers, but that's the link to watch.

Then there's supply. AMD plans to "substantially increase" it in 2027, which is good for sales until everyone else does the same. China offers a warning: the STAR50, an index of mainland AI leaders, has dropped more than 40% since its summer peak. One recent analysis frames that slump as a warning and asks what happens when the supply of new tech catches up with demand.

So here's what I'd watch: whether AMD's margins hold as its 2027 supply arrives, whether Intel's shortage eases, and whether AI borrowers keep getting funded at tolerable rates. If margins hold while supply grows, that's good evidence the demand is real. If they crack, this quarter may turn out to be what peak shortage looks like.