Discussion about this post

User's avatar
James Emanuel's avatar

At the recent earnings call, Su was asked about whether we are in an AI bubble.

Her response was not what I had expected.

In an effort to downplay the bubble theory, she stated that at times like these it is better to overspend than to underspend.

Does that not reinforce the bubble theory?

Is that not reckless for a CEO? How can she advocate overspending which implies a misallocation of shareholder capital?

Could this be her first major mis-step?

I welcome your views.

Thesis Rationale's avatar

Hi Neural Foundry, Absolutely, this is a great point, and I fully agree with your framing.

The industry often talks about Zen’s IPC leap as the headline breakthrough, but the real structural unlock was chiplet modularity. It let AMD break free from the yield, cost, and scheduling bottlenecks that constrained monolithic designs for a decade. Once AMD shifted to a distributed architecture, iteration speed increased, binning economics improved, and performance scaling became fundamentally more predictable.

And you’re right about the TSMC dependency. People label it as a risk, but in practice it was a strategic liberation. Instead of being slowed by Intel’s in-house process delays, AMD plugged directly into TSMC’s leading-edge roadmaps, by getting access to 7nm, 5nm, and eventually 3nm ahead of most of the industry. That alignment allowed AMD’s architectural progress to compound without fabrication uncertainty dragging it backward.

In short: Zen made AMD competitive again, but chiplets + TSMC are what made the turnaround sustainable.

Thank you for your insight!

3 more comments...

Ready for more?