korrents

korrents · video

Dylan Patel – Two labs will soon control most of the world's workforce

Dylan Patel · 1h 16m · youtube.com ↗

30 korrents from this recording

1h
Dylan Patel did not write this page.

Every claim below is a statement made in this recording, quoted word for word and linked to the second it was said, so you can hear it rather than take our word for it. The wording comes from the transcript published alongside the recording; the sentence above each quote is our reading of the claim, not their wording.

  1. 0:00:36 · watch on youtube.com↗

    Yeah, so when we go back to last year, even at the end of the year, most of GDP growth in America was just AI infrastructure.
  2. 0:01:03 · watch on youtube.com↗

    We're at a little bit over a trillion dollars of CapEx this year. As we go out into 28, it's going to be more than $2 trillion.
  3. 2 min later
  4. 0:02:50 · watch on youtube.com↗

    But now, when OpenAI serves GPT 5 .6 or Anthropic serves Opus 5 or Mythos, Fable 5, their revenue generation has passed well beyond the sort of incremental $10, $15 million per megawatt.
  5. 2 min later
  6. 0:04:24 · watch on youtube.com↗

    And this centralization doesn't look like it's slowing down or stopping. In fact, it looks like it's only accelerating.
  7. 2 min later
  8. 0:06:47 · watch on youtube.com↗

    So by the time you're in, like, towards the end of 2028, if this trend continues, which I see nothing that's stopping it, you've got them just controlling most of the usable, you know, flops in the world
  9. 4 min later
  10. 0:10:28 · watch on youtube.com↗

    But ultimately, yes, capitalism will cause these things to expand, but it's a whip, right? It takes a long time for the whip signal to get to the tail end of that. And so the supply chain doesn't react immediately.
  11. 2 min later
  12. 0:12:50 · watch on youtube.com↗

    You sum all this up, it's going to be well north of $2 trillion of CapEx. So the labs have not yet gotten to the point where their cash flows can fund this stuff.
  13. 4 min later
  14. 0:16:21 · watch on youtube.com↗

    This regulation that they advocate for is actually slowing down the labs a lot more than it slows down, you know, sort of the open -source Chinese language models.
  15. 0:16:41 · watch on youtube.com↗

    They're clearly not releasing their best models. And in which case, their revenue per megawatt stalls or even can start to decline again because other models are competitive again.
  16. 2 min later
  17. 0:18:29 · watch on youtube.com↗

    Like most of the value that these models generate does not get given to OpenAI Anthropic. Thankfully, so far, it is mostly just being given to the users, right?
  18. 10 min later
  19. 0:28:49 · watch on youtube.com↗

    more and more the method of regulation is new york's banning data centers texas is holding memoratoriums ohio saying you have to or at least trying to say you have to like pay everyone's property tax in a certain radius these sorts of things are going to decrease supply and increase cost
  20. 0:29:13 · watch on youtube.com↗

    in a takeoff scenario why would anthropic not have their best model six months ahead of what is externally available because of safety and regulation but also you know the competitive advantage
  21. 1 min later
  22. 0:30:30 · watch on youtube.com↗

    I personally believe that the labs are going to allocate less and less compute to inference over time, which I think is very non -consensus, right?
  23. 1 min later
  24. 0:31:10 · watch on youtube.com↗

    Or do you go build AGI? And I think the obvious answer from Anthropic and OpenAI, and not just at the executive level, but also their board, is go build AGI because it's way more profitable.
  25. 1 min later
  26. 0:32:15 · watch on youtube.com↗

    ultimately it makes more sense to dedicate more and more compute internally. And the only reason to have inference compute be so large is so you can grow your training fleet.
  27. 3 min later
  28. 0:34:58 · watch on youtube.com↗

    So today, 70 % of watts are being deployed in America. And, you know, China is really a very small number. It's sub 10 % of watts being deployed for data center AI compute is in China.
  29. 1 min later
  30. 0:35:33 · watch on youtube.com↗

    So in 2028, it might start to inflect up, I think. But it's pretty easy to say China will have like 30 gigawatts of AI compute or less.
  31. 2 min later
  32. 0:37:17 · watch on youtube.com↗

    But ultimately, you know, China is definitely going to hockey stick if there's anything China's really good at is scaling manufacturing really, really quickly.
  33. 1 min later
  34. 0:38:32 · watch on youtube.com↗

    Whereas China is not going to slow down AI. In fact, the only thing they're going to do is accelerate it.
  35. 1 min later
  36. 0:39:46 · watch on youtube.com↗

    the Chinese semiconductor industry has significantly more subsidies than the rest of the world's semiconductor industries combined
  37. 0:40:09 · watch on youtube.com↗

    Chinese companies today are not that far behind in AI models, at least perceivably by the public, relative to the amount of compute they have, right?
  38. 1 min later
  39. 0:41:03 · watch on youtube.com↗

    And that is actually, like, 50 % of the compute is research, like, 10 % of the compute is development, and then 40 % is inference.
  40. 1 min later
  41. 0:42:25 · watch on youtube.com↗

    implement automated coding, automated researcher, I actually expect the percentage of the compute budget that goes to research versus training to start to, like, become a lot more fuzzy or even higher for training.
  42. 3 min later
  43. 0:45:40 · watch on youtube.com↗

    But they now don't generate cash. They actually spend everything on CapEx. And in addition, they raise debt and spend everything on CapEx, right?
  44. 8 min later
  45. 0:53:55 · watch on youtube.com↗

    So, all of these things bend the curve from what does capitalism want in terms of just pure, simple economics to what is the complex system that we have want and bends it lower and lower and lower to where not as many gigawatts as should be built will be built.
  46. 1 min later
  47. 0:55:21 · watch on youtube.com↗

    No, I'm just saying until 2029, there's, you know, something on the order of $11 trillion of CapEx. And six of that is funded with cash. And five of that is funded with debt.
  48. 7 min later
  49. 1:02:45 · watch on youtube.com↗

    the limiter on agi is not how fast can the research engineers like you know our roommate sholto can crank the gears it's actually just like how much does the rest of the world let that happen right
  50. 8 min later
  51. 1:10:59 · watch on youtube.com↗

    Because it seems to me that every force is screeching toward centralization. Right. And that's scary as hell.
  52. 2 min later
  53. 1:12:47 · watch on youtube.com↗

    I don't trust the government, and I don't trust Dario, and I don't trust Sam.
  54. 2 min later
  55. 1:14:29 · watch on youtube.com↗

    everything along the way will still lead to someone's going to allocate, going to capture more resources. So, it's kind of hard for a framework in which AI doesn't lead to super concentration.