Skip to playerSkip to main content
  • 20 hours ago
Transcript
00:00There are some details worth understanding and knowing about in what Anthropic plans to do with
00:04this $50 billion over which timeline and with who. But there's also some consideration around how big
00:11this $50 billion is relative to the rest of the field. That's right. You know, I think in one
00:16sense, if you had said a year or two ago that a four-year-old unprofitable startup was going to
00:21commit $50 billion to an infrastructure project, you'd say that's enormous and possibly insane.
00:26But when you compare it to the $1.4 trillion that OpenAI has committed, it feels much more measured.
00:32And I do think that is very much the subtext of a lot of this. Anthropic is clearly taking a page
00:37from OpenAI in terms of the scale of infrastructure required to build what it wants to build. But I
00:42think it also wants to show that it's perhaps a more fiscally responsible company in doing this
00:47in order of magnitude less at the moment. They've raised $13 billion recently. They're worth $183
00:53billion. And they've had really good track record with business and enterprise sales, 300,000 business
00:59customers. But what's interesting is the business they're doing with other people. They're using a
01:03UK company to build out in the US, Seth. That's right. I mean, FluidStack is really not a household
01:07name here. I mean, it's part of that emerging sector of neoclouds, but not nearly as well known as
01:12CoreWeave or even Nebius. But, you know, it's an up-and-coming company that we've previously reported
01:16years and talks for a large amount of funding. And it's been central to a French effort to build
01:21a supercomputer over there. But again, more untested perhaps than some other partners you might
01:26announce here. It's one of the stories where it's what we do not know that's worth consideration.
01:33So they're not telling us, you know, the total gigawatt figure for the footprint in the United
01:37States. And at the same time, you know, they're heavily reliant on these hyperscale partners, Amazon, AWS,
01:45and Google, to give them the compute they need away from this plan to build $50 billion of capacity.
01:51Yeah. And I think, again, if you look at their nearest rival, OpenAI, the answer is sort of it takes
01:55everything and everyone to build what they want to build here. And so while they want to potentially have a bit more
02:01control and ownership over some of the data centers that they build in the coming years, they're still going to be
02:05relying on those big tech partners, and the neoclouds and pretty much anyone else in this industry to meet what
02:11they imagine to be a pretty intense amount of computing needs here.
Be the first to comment
Add your comment

Recommended