The chipmaker everyone crowned emperor just got outrun by the scrappy underdog harder than any time in a quarter century, and now earnings will tell us if it's real or just noise.

The Summary

The Signal

AMD just posted its best five-day performance against Nvidia since before the dot-com bubble burst. The 53.2 percentage point spread is not a rounding error or a bad trading day. It is the kind of divergence that happens when money starts questioning a narrative it used to treat as gospel.

For two years, Nvidia has been the default answer to "who wins AI infrastructure." Its H100 and H200 GPUs became the compute layer for everything from ChatGPT to Midjourney. AMD was the other guy, the one you considered if you couldn't get Nvidia allocation or if you were building something weird. That positioning just shifted hard enough to show up in charts from 1999.

"The significant market shift suggests evolving investor confidence in AMD's growth potential, challenging Nvidia's dominance in AI hardware."

What changed? AMD's MI300 series started landing real contracts. Hyperscalers are not just kicking tires anymore. They are deploying at scale, and they are doing it because AMD chips cost less and integrate better with existing infrastructure for certain workloads. The AI compute market is not winner-take-all. It is fragmenting by use case, and AMD is carving out territory where Nvidia's premium does not make sense.

But then Friday happened. AMD dropped 8.75% while the broader market stayed flat. That is profit-taking, sure, but it is also doubt. The rally was fast and sharp, the kind that attracts momentum traders who bail at the first wobble. The real test is not whether AMD can spike relative to Nvidia. It is whether it can hold gains when Nvidia posts numbers.

  • Nvidia reports earnings November 17
  • If Nvidia beats and guides up, does AMD's rally evaporate?
  • If Nvidia disappoints, does that validate AMD's shift or just tank both stocks?

The Kobeissi Letter's chart going back to 1999 is not just trivia. It frames this moment against Nvidia's entire public market history. The last time AMD had this kind of relative momentum, the internet was still on dial-up and crypto did not exist. The comparison to Cisco, implied in The Pomp Letter's framing, is darker. Cisco was the infrastructure king of Web1 until it wasn't. Its stock never recovered from 2000. The question hovering over Nvidia now is whether AI infrastructure follows the same script: explosive growth, market dominance, then commoditization and collapse.

The Implication

Watch what happens after Nvidia reports on November 17. If Nvidia beats and AMD holds its gains, that confirms a real shift in how the market values AI compute. If AMD gives it all back, this was just a trade, not a trend. For builders in the agent economy, the takeaway is practical: AMD is now a credible alternative for inference workloads and multi-modal training at scale. That means better pricing power and more leverage when negotiating compute contracts.

For investors, the Cisco comparison matters. Nvidia's valuation assumes it stays dominant and high-margin forever. If AMD proves there's a viable challenger, that assumption breaks. The infrastructure layer commoditizes. Margins compress. The profits move up the stack to the companies building agents, not the ones selling shovels.

Sources

BeInCrypto | Crypto Briefing | The Pomp Letter