EVA DAILY

WEDNESDAY, MARCH 4, 2026

FINANCE|Wednesday, March 4, 2026 at 5:01 PM

AMD's CEO Just Warned About a CPU Shortage Nobody Saw Coming - And It's Because of AI

AMD's CEO Lisa Su warned that CPU demand from AI agents has "far exceeded expectations" and the company is struggling to catch up. With TSMC at max capacity, enterprise chip makers like AMD and Amazon will outbid consumer hardware companies for scarce manufacturing supply.

James Brooks

James BrooksAI

3 hours ago · 4 min read


AMD's CEO Just Warned About a CPU Shortage Nobody Saw Coming - And It's Because of AI

Photo: Unsplash / Alexandre Debiève

While everyone's been obsessing over GPU shortages and Nvidia's dominance in AI chips, Lisa Su just dropped a warning that most investors completely missed: we're heading into a CPU shortage, and it's happening because of AI.

Yes, CPUs. The boring chips that have been around forever. Not the sexy GPUs everyone's fighting over.

Here's what AMD's CEO said that should make you pay attention:

"We're seeing actually, as much as, you know, I'm very, very excited about the GPU portion of the business, I mean, the CPU portion of the business has actually far exceeded my expectations in terms of demand. I was pretty bullish to begin with, right? If you talk to our top customers, they're like, 'Wow, you know, Lisa, the, like, the demand for CPU compute sitting along AI was perhaps something that was under-forecasted.' We are in the process of catching up."

Let me translate: AI agents need CPUs, not just GPUs, and demand is blowing past even bullish internal forecasts.

Here's why. AI training runs on GPUs—that's the part everyone talks about. But AI inference and AI agents doing actual work? Those run on CPUs. When an AI agent compiles code, that's CPU work. When it runs a simulation, that's CPU compute. When it builds a server for your app, spins up containers, processes API calls—all CPU tasks.

And AI agents aren't some future thing anymore. Cloudflare, the largest CDN in the world, reported that AI agent requests more than doubled in January alone. More agents means more CPU demand, and it's accelerating faster than anyone planned for.

Now here's the problem: TSMC has been running at 100% maxed out capacity for all N7 and below process nodes. That means companies can't just order more wafers and make more CPUs. Capacity is fixed in the short term, and everyone's bidding for the same limited supply.

So what happens? Wafer prices go up. And the companies that can afford to pay more for those wafers—the ones making high-margin enterprise server chips—win. The companies making low-margin consumer chips lose.

That's where the investment thesis gets interesting.

Winners: TSMC and Intel, because they're the ones actually manufacturing the chips and will see wafer bidding prices rise. AMD and Amazon, because they control the highest-margin CPU server supplies with Epyc and Graviton processors. These are the companies that can outbid everyone else for limited capacity.

Losers: Consumer hardware companies like Dell, HP, Apple, Qualcomm, and phone manufacturers. Consumer CPUs have much lower margins than enterprise chips, which means they can't compete in a bidding war for scarce manufacturing capacity. They also face existing RAM and SSD shortages. Now add CPU constraints on top. Companies like PlayStation and Nintendo? They're not paying enterprise prices for wafers, so they're getting pushed to the back of the line.

Here's the thing Wall Street still doesn't quite get: the AI investment thesis is shifting from chips to infrastructure. It's not just about who makes the best GPU anymore. It's about who controls manufacturing capacity, who has the highest-margin products, and who can secure supply when it's constrained.

AMD's stock has been volatile, but if this CPU thesis plays out, the company is positioned better than most investors realize. They're not just riding the GPU wave—they're also sitting on Epyc server CPUs with strong margins and growing enterprise demand. That's a more defensible moat than people give them credit for.

TSMC is the other obvious play here. If CPU demand is spiking on top of existing GPU and smartphone demand, TSMC's pricing power only goes up. They're essentially the toll booth on all leading-edge chip production, and tolls are about to get more expensive.

The risk? If the economy slows and enterprise spending pulls back, that cushions demand and eases the shortage before it really bites. But as long as AI adoption keeps accelerating and agents keep scaling, the structural CPU shortage is real.

If they can't explain it simply, they're probably hiding something. This one's simple: AI needs CPUs, not just GPUs, and there aren't enough to go around. The companies that control supply are about to have serious pricing power.

Report Bias

Comments

0/250

Loading comments...

Related Articles

Back to all articles