Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Alrighty guys. So now that we've had a couple of
(00:02):
days to really dig into this deep seek thing and
the effects of the stock market, I wrote a couple
notes down because I've been doing some research on this,
because you guys know, I'm deep in the AI world,
and so this is something that I think actually is
a positive thing. So I'll just give the TLDR now
the whole Deep Seek all of this happening. It's a
(00:24):
good thing. Let's get to it. The last couple of years,
we've gone through an AI boom, that's clear. So much
money has gone into the AI world from venture capital,
private equity, investment banks, et cetera, et cetera. So what
(00:44):
we've had essentially is this AI bubble that has grown.
And some might say, is deep Seek going to pop
the A bubble? No, the answer is no, it's actually
popping something else. And so what we've found through you know,
deep Zeek is that currently like chat gibt, it's like
(01:05):
two dollars and fifty cents per million tokens right to
be able to run the run the systems. And what
we're finding is that deep Seek is basically zero point
to two seven basically twenty five cents per million tokens, right,
and so that's like nine point seven times more effective.
Like rounding up, it's an entire order of magnitude cheaper
(01:30):
to run at the top level. And so what we've
seen is this massive amount of spending and capital expenditure
into AI, AI tech hardware right in VideA is just
that's where a lot of the bloat is right. And
so what we have found and this is the This
(01:51):
is the end of the story, guys, not a long video.
The end of the story is the AI bubble has
not popped. What I'm thinking is that the capital expenditure
bubble into AI has popped with an entire order of
magnitude differential. Here, what this means on the opposite side
(02:14):
is that it is actually ten times cheaper than we
thought to run these models. This is a good thing.
If it's ten times cheaper to run the models, then
the big players your Facebook, your metas, right Chat, get Grock,
write all that, right, they will continue to move forward
(02:37):
in AI. But with this new revelation, there are clearly
ways to run it exponentially cheaper. And this is also
good for the overall market. This means that the smaller
players and the medium sized players. The startups actually have
less capital expenditure needs to be able to run and
rapidly prototype their ideas. This for me as a venture
(03:00):
capitalists is juicy right because as the startups find opportunities
to be able to run and build their applications at
an order of magnitude cheaper, we might actually see an
application boom, like if you think about it, on the
(03:20):
second layer, I think we're going to be signed o