Real Alpha
Try Pro free

Prof G Markets · 3 min · Oct 26, 2026

AI chip delays may reflect power bottlenecks, not weak demand

Scott Galloway

Scott Galloway says idle GPUs can reflect power and data-center bottlenecks rather than weak demand.

In this episode

Full analysis
  1. 7:58The merger shifts attention from Tesla’s weakening car businessPro pick 8 min
  2. 18:43A SpaceX merger could dilute Tesla shareholders’ protectionsPro pick 5 min
  3. 55:55Public markets price AI risk more cautiously than private markets5 min
  4. 34:04Merger committees should test premiums against alternatives5 min
  5. 51:03AI chip delays may reflect power bottlenecks, not weak demand3 min
  6. 1:08:45Meta's filing suggests AI funding could become harder4 min
“the bottleneck has moved from chips to electrons.”
Scott Gallowayat 51:21 · Prof G Markets
51:21

Takeaways

  1. Amazon's GPU backlog may reflect a power bottleneck.
  2. Data-center construction delays can leave hardware waiting to be installed.
  3. Scott sees correction risk but not an imminent implosion.

About this segment

Scott Galloway says idle GPUs can reflect power and data-center bottlenecks rather than weak demand. He sees a likely correction, not an AI-market implosion.

Scott Galloway, Host

From “The $3.5T Bailout: Elon's Plan To Merge Tesla And SpaceX”, Oct 26, 2026. Watch the full episode

Answers “Is AI a bubble?” See all 30 answers

Answers “Is there still a GPU shortage?” See all 35 answers

Answers “Is there enough power for AI?” See all 29 answers

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

  1. 51:16

    too much time with Paul Kedrosky and Ed Citrin. I think we're in the midst of a correction, but not an implosion. And that is, so first off, with respect to the GPU stacking up in Amazon, which is arguably the best supply chain operator in history, the bottleneck has moved from chips to electrons. And so the idea that when there seems to be the scarce resources chips and Amazon uses its muscle to stockpile them, thinking that's a smart move, and then we find that the bottleneck is around either power or data centers themselves, I don't think that's that unusual a story.

  2. 51:55

    And media has, media narratives have a momentum. Everyone gets in favor of something or you can just feel the momentum around what is the narrative. And the narrative right now is that the catbacks in AI is way out over its skis and there's no way they can get an ROI there and we're going to see a 99 or a 2000 like implosion. and part of that narrative i've been reading is that okay uh anthropic if you look at their numbers the media is saying all right on 16 billion in revenues they lost 80 billion all right that sounds very scary but what they fail to say is that 72 of that 80 billion dollar loss was the loss the accounting loss they have to take because of convertible debt that converts into equity.

  3. 52:39

    because the equity keeps going up in value. So really the kind of accurate number is they lost $9 billion on $16 billion in revenue. And I think they 9X'd their revenue year on year or something. So I do think these companies, the AI trade or the technology is still showing an unprecedented level of velocity, both in terms of revenues and adoption and new products and capital formation. Now, is that not to say there won't be a correction?

  4. 53:12

    I agree there's going to be a correction here. But right now, the narrative is kind of turned into this is so clearly a bubble waiting to pop viciously. And I'm just not sure I buy that. The neoclouds, Nebius is down 26 percent, Corio is down 27, Cypher Digital is off by half.