00:00:00
To the extent that open AI missed, I
00:00:03
think what that is is an insight to not
00:00:05
enough compute capacity today. And that
00:00:07
problem is only getting worse. Okay, if
00:00:09
you look at the actual amount of
00:00:12
gigawatts that are under construction,
00:00:14
we have a huge mismatch now. People have
00:00:17
announced all these projects, but less
00:00:19
than half of it is actually being built.
00:00:21
Less than half. Most of it is stuck in
00:00:24
red tape. There's no credible strategy
00:00:27
to turn any of this stuff on.
00:00:29
Who will this hurt? It will hurt
00:00:32
Anthropic and OpenAI the most. Who will
00:00:35
this benefit? It will benefit the
00:00:37
hyperscalers, specifically Oracle,
00:00:39
Amazon, Meta, Microsoft, and Google. And
00:00:42
now what you're going to see is a
00:00:45
negotiation and a trade back and forth.
00:00:47
How much equity do I have to give up?
00:00:50
How much control do I have to give up to
00:00:51
get access to the compute versus how
00:00:54
badly will I miss my growth forecasts if
00:00:57
I don't? That's a huge lane for Grock to
00:01:00
just run through and SpaceX to run
00:01:01
through because they have a ton of
00:01:04
excess capacity. And so I think the
00:01:06
cursor deal was the appetizer. If I were
00:01:09
Elon now, I'd be running all over this
00:01:11
market because if the models catch up in
00:01:13
quality, I think he could also do
00:01:16
something really crazy with anthropic.
00:01:18
He and Dario should do a deal