What's Next For U.S. - China Artificial Intelligence Rivalry — Saagar Enjeti

Channel: Alex Kantrowitz

Published at: 2025-02-20

YouTube video id: Xz63EixPD_w

Source: https://www.youtube.com/watch?v=Xz63EixPD_w

in the way that the Clinton presidency
is largely remembered for its transition
to the internet I think that in the long
term we may view the Trump years as the
second Administration as the AI
presidency in terms of the way that our
economy and our life and all of that
were transformed and in that I think
there's a lot going on here and um first
and foremost is an interesting Clash
within technology itself so this will be
a huge fight going forward around export
control so a lot of people were passing
around uh after the Deep seek Revelation
but that Alexander Wang clip where he
was like look they obviously have more
Nvidia h100s than they can admit because
they're going around export controls and
then I was not even aware of this how
20% or whatever of Nvidia revenue is in
Singapore I'm like yeah I'm sure
Singapore leads a lot of chips it's like
oh fascinating right it's obviously
leakage in terms of uh export control to
China it makes sense too why Nvidia was
pushing back so hard in the later days
of the Biden Administration against
tightened export control rules I hadn't
followed it fully at the time but
putting it all together I'm starting to
it's starting to click um how much of
Revenue that they alone are also kind of
relying on by a trans shipping chips But
but so at a broader level we have like a
couple of choices and interesting paths
like one would be okay we should just go
full open source we we should the these
companies should stop individually
chasing this 75 billion per quarter on
AI we should focus on efficiency well
one of the Tells to me that deep seek
was not quote unquote as revolutionary
as people said is because the companies
are still spending the money and so what
they took away from that as I understand
by open Ai and others is that they saw
that their model was used by the Deep
seek hedge fund to train off of and so
actually what's happening is the net
effect of what happens whenever you have
kind of breakthrough technology if we
think about it like the atom bomb um
well what happens is that inevitably you
have scientists copying other scientists
and so that leads to closed circles
secrecy closed models National Security
concerns and overall like geopolitical
uh geopolitical uh implications of your
individual business decisions such that
it no longer really becomes about Nvidia
or meta stock it becomes about the
United States of America and I think
this will be a very difficult thing for
NVIDIA for meta for Google and all of
them to deal with because I think that
in their hearts at least in the past
they really were globalists and it makes
sense you know meta is they're user
based the vast majority of them are not
Americans but your headquartered here
sorry and Mark Zuckerberg you're waving
your American flag around on your
surfboard or whatever so you've you've
planted your flag L quite literally how
far we've come from trying to name his
kid after she she's yeah that's right
yeah after doing that as you and I know
which is Hil even more hilarious in
retrospect so putting all of that
together with the Deep seek thing you
basically see it laid out by JD Vance in
his policy speech where he's like we
want America to be the leader in AI we
will not allow you know Espionage on our
companies so we're going to see literal
walls begin to go up much tighter export
control a a recognition that these
companies and this technology is vital
and of interest and uh I I don't really
think there's any way out of it to be
honest because you know I was really
thinking about it and it's like well
what is the alternative because we could
try to have some path of not going down
that but it's very clear that the
Chinese um especially in the case of
deep seek I mean this idea that it's $6
million or whatever to run is ludicrous
it obviously cost over a billion dollars
uh to train the model before it
eventually got to the point where it
could run off of that 6 million it's
clearly subsidized and at least pushed
and allowed um by the Chinese government
who themselves I'm not sure you've seen
Alex but the Chinese Tech SCS tech
stocks are roaring um on the Hong Kong
Stock Exchange right now uh as a result
of the deep seek thing so the government
has its own interest and control and all
that behind the scenes for them and the
Chinese are never going to move away
from their own firewall model so in a
sense you know naturally the United
States especially to preserve so much of
the Enterprise Value and the genu
technological breakthrough doesn't
really have a choice but to behave quite
similarly and I just think that's an
inevitable kind of FUSD strap of any
great superpower competition on any
technological basis I want to ask you
the a similar question that I asked
Alexander Wing when he was here which is
um why is it such a strategic liability
if China takes the lead on AI models
like just take deep seek for example yes
they definitely trained with a lot more
money than they said they did to build
this R1 model however the R1 model is
like undeniably cheaper to run and so
now you have us startups using that
model to build products they probably
couldn't build in the past so talk a
little bit about why it's so
strategically important for the us to
have the lead and not not China there's
a couple different ways I mean uh first
and foremost is just um well actually
let me give you a really meta way that
he's probably never going to admit and
this is the truth America has decided
that our retirement plan is number go up
and our numbers need to go up that
number can't go up if we're number two
and if we're subject uh in the 20% of
our GDP which is technology the fastest
growing sector uh the Magnificent 7
comprises what percentage of the S&P 500
like 20 to 25% depending on the day 30%
sometimes let's let's look at this
realistically like at a basic level we
don't have a choice like we have all
chosen that America retires because the
number goes up and the number's not
going to go up if we're not the leaders
in the technology because it means that
we won't be able to unlock the
Enterprise Value and have all you know
all of the benefits of network
capitalism and AI but at a more like
basic level the United States especially
with respect to technology with uh it's
with research and everything the
implicit you know deal that we've all
made is that like I said we're not going
to have government retirement or
government Healthcare or any of that but
in exchange we will be the leaders of
the global economy of the world um and
you know rule the rules-based
international order to the benefit of
the American Consumer and so at a really
like high level strategic wise it's the
same reason why Sputnik simply could not
be allowed to stand it's because even if
the terms and the way that we're
fighting seem ridiculous the notion of
space Supremacy by our adversary or
technological Supremacy beating us in
that category just means their ability
to compel our Behavior so at a very like
meta level in terms of why it's
strategically important it really is
gets back to that thusi trap idea of we
don't when your adversary is to choose
development down a closed path which
would allow technological dominance over
you you either have a choice to like
reach your hand out shake hands have
some mutual agreement where that's
simply not going to happen and then have
the trust the enforceability in such a
contract that you can ensure things stay
smooth that's not going to happen
unfortunately or you have the chance to
respond and so uh yeah I just I don't
even think of it as AI I think about it
as the way I would any breakthrough
technology I would give you the exact
same answer yeah and there's been some
talk that like okay well AI is both a
business uh but it's also a weapon and I
am curious like you know JD Vance
recently said the fears about AI are
overblown so where do you stand on that
question because I I think about it
every day so every day I either am an
alarmist or I just think it's all
ridiculous so you and I it's funny how
you go back and forth same thing happens
to me where I'm like this chat GPT can't
tie its shoes but on the other hand
you're like oh no this can control
computers and build deep research
reports like oh God what's going to
happen next that's how I feel so I use
CLA I'm a claw guy I can't really tell
you why I just like the interface um for
data anal we talk about it a lot it's
pretty nice to talk to like actually
feels less botlike yeah yeah I totally
agree and you know for data data
analysis look I'm not the typical person
I'm not a coder any of this stuff you
know what I really use it for personal
finance I go how much money have I spent
on groceries this year and I'll be like
here this and this is your break that's
a pretty cool use case I haven't heard
that before so you like upload your like
bank statements and talk to I have a
budget app which has all my trans
actions and then I downloaded my
transactions for the last seven years
uploaded it to Claude And I go so how
much money do I spend on milk how much
money do I spend on eggs um what's my
average burn rate on travel so and then
also you can uh you can calculate to the
literal dollar not only your spending
but then your not your discretionary
spending your non-discretionary spending
and then if you're trying to budget and
project like let's say Investments I use
it for investment analysis all the time
so I'll be like if I dollar cost average
$200 a month into x stock based on
historical Returns what is my expected
over a 10year period And it it's spits
out charts it's fantastic should I have
it do my taxes like I'm thinking like I
can just upload all my books and see
what I know the problem is legally
Legally you're it's you're getting dicey
right so like when I'm doing some
stock analysis like the stakes
are not that high but with doing your
taxes I I would
obviously yeah I mean maybe you should
try it tell me tell me what you think
keep you posted case thinking about AI
um do you have you ever used Tesla full
self-driving I haven't driven with it
but I've been in the car when someone
put it on okay so uh I have a Tesla and
I have full self-driving so that's where
I don't buy the AI hype I have been
hearing for years oh this car is going
to drive itself no it's not Tesla can
barely turn left bro and even when it
does even when it does turn left it
turns left like a right angle you know
or uh just the other day you'll love
this example cuz this just shows how
Dynamic and difficult it is to try and
accomplish this problem it snowed and
the snow got caught in the lights the
red light like in the the cup that holds
the lights so the Tesla could not
distinguish whether it was red or
whether it was green so my eye could I
could tell like oh right up there
there's a little bit of red light green
light but there's so many different
things or there's areas where I live
where you can park the in this Lane
sometimes but not others and so the
Tesla if it sees it open will want to
get into that lane even though I know
you know like oh it's 400 p.m. I can't
get there my point is is that I've even
with all of the hype and even with let's
be real the best self-driving technology
available in the US it's not that useful
unless you're on the highway is it that
much better than Toyota's assist or than
Honda's assisted uh assisted autopilot
yeah it's good it can change lanes and
stuff like that but I've been in a BMW
that can change its own Lanes so my
point is is that even with Tesla which
has thrown billions at this problem
which is the world's leader and all of
that I've owned this thing for a couple
of years and it's definitely gotten
better but it's not that great and
that's as in I'm not I'm not taking my
hands off the wheel anytime soon