June 14, 2025 • 26 mins
Back in May, President Trump announced this big plan whereby American tech giants would participate in major AI projects in both Saudi Arabia and the United Arab Emirates. The announcement has created divisions within the White House, and more generally among people who are thinking about the intersection of artificial intelligence and geopolitics. One argument is that this is great news geopolitically, because it gives American technology a beachhead in this crucial region. Another argument is that by exporting the chips abroad, it creates a possibility that some of the technology will leak to China, or benefit China in some way. On this episode we speak with Bloomberg News reporter Mackenzie Hawkins, who covers tech and geopolitics in Hong Kong, about the deals, the divisions, and what to watch next as the US looks to maintain its edge in these key areas.
Only Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox now delivered every weekday plus unlimited access to the site and app. Subscribe at bloomberg.com/subscriptions/oddlots
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:02):
Bloomberg Audio Studios, Podcasts, radio News.
Speaker 2 (00:18):
Hello and welcome to another episode of the Odd Lots podcast.
Speaker 3 (00:21):
I'm Joe Wisenthal and I'm Tracy Alloway.
Speaker 2 (00:24):
It's a few weeks ago, but last month Trump was
in the was in Saudi Arabia, he was in the
Middle East, and he sold a bunch of chips.
Speaker 3 (00:30):
Yeah, I think the chip's announcement was kind of overshadowed
by some of the pomp and ceremony around that visit.
Did you see the famous hair flipping No, oh, you
didn't see that, So you need to look this up.
So when Trump arrived in the UAE, part of his
big you know, the greeting and the red carpet getting
rolled out for him was a bunch of emmiati women
(00:52):
doing a hair flipping dance. I gotta say, I wish
I could flip my hair like that, but sadly I
could not.
Speaker 2 (00:59):
Can I say, like Trump seems very at home in
the Gulf States, you know what I'm saying, Like it
feels like, setting aside many other things that there is
like a similar sense of esthetics.
Speaker 3 (01:11):
Grant I was about to say that he's like, I think.
Speaker 2 (01:14):
He's very comfortable in those countries.
Speaker 3 (01:16):
I was about to say, his taste in interior decorating
is exactly the same as what you see in palaces
and hotels in Saudi Arabia and places like that.
Speaker 2 (01:25):
Absolutely, No, I think just the whole vibe is like, oh,
his on home territory. But I think the AI stuff
is really interesting. Obviously, for a lot of reasons, the
US wants to sell more things. One of the things
that we can sell in various flavors is AI technology.
But there's this weird like I don't know what the
word cross currents or whatever, which is that we want
to sell more stuff, and yet we also don't want
(01:48):
our stuff to be in the hands of perceived geopolitical rivals,
and so we cut off our own sales. China is
obviously the most salient example here, but it sort of
makes this interesting tension when we think of the role
of AI sales tech sales in sort of international relations
and diplomacy.
Speaker 4 (02:04):
Yeah.
Speaker 3 (02:04):
Well, I also find this interesting from the UAE slash
Abu Dhabi side of things, right, because the UAE, it's
strategically aligned with the US in many ways, but it's
very commercially aligned with China, and you know, you have
a lot of Chinese commerce coming through the ports and
things like that, so I'm very curious what they see
(02:26):
in this steal and how they're navigating those two sides
as well.
Speaker 1 (02:29):
Well.
Speaker 2 (02:29):
I'm very excited we have the perfect guest, someone one
of our colleagues we've had on the podcast before. We're
going to be speaking with Mackenzie Hawkins, tech and geopolitics
reporter Bloomberg News. She's now based in Hong Kong, sadly
for us, but thankfully she's you know, staying up late
for this recording. Mackenzie, thank you so much for coming
back on odd locks.
Speaker 4 (02:49):
Thanks so much for having me.
Speaker 2 (02:50):
First of all, why don't you just give us the
specific details, as the salient details as you understand them
in terms of this big, big AI deal that Trump
announced when he did his golf tour.
Speaker 4 (03:06):
So in the four day span or so that Trump
was in the Middle East, you saw American tech companies
and companies in golf nations announce hundreds of billions of
dollars worth of AI deals. Specifically, many of them focused
on AI infrastructure, building out the data centers that house
AI chips that are used to train and deploy large
language models, and that involved one of the biggest data
(03:30):
centers potentially in the world, a project anchored by the
Abu Dhabi AI firm G forty two, with key participation
from Oracle, from Open AI, from Nvidia, huge volumes of
Nvidia ship sales going to both the UAE and Saudi Arabia,
as well as sales from Nvidia's rival Advanced micro Devices,
and all told, we sort of ended the week, and
(03:53):
the deals announcements actually continued into the following week. We
ended this trip going from a world where the two
main players with significant AI data center capacity the US,
of course, and then China also has major ambitions and still,
even if all of these deals fall through as expected,
which is an if, the US will maintain the vast
(04:15):
majority of the world's computational power, but we have two
new serious players in the UAE and in Saudi Arabia.
Speaker 3 (04:23):
One thing I know from living in Apudapi for a
couple of years is that Apudapi loves building a giant
campus with some sort of foreign partner like that is
always happening and seems to be happening more and more.
But okay, so from the US side. You know, the
US wants to be a powerful player in AI and
everything that comes with it, chips, data centers, et cetera.
(04:44):
And they want to sell that technology to other countries.
What's the issue with the deal? Why is it controversial?
Speaker 4 (04:51):
So there are a couple of sort of categories of
disagreement in Washington, if you The first is a basic
economic competitive this argument. You have people, especially on the
Democratic side of the aisle, saying, you know, why are
we building a data center in Abu Dhabi when we
could be doing that in Ohio or in Pennsylvania. The
response from people who are in favor of these deals is,
(05:12):
we are building data centers in the US. Trump is
trying to cut red tape to facilitate data center deals
across the country, and the UAE accord, in particular, because
there's now a bilateral arrangement between American and AMRADI officials,
includes this reciprocal investment portion where you know, for the
amount that US companies are investing in the Gulf nation,
we'll get a reciprocal amount on American soil. But there
(05:35):
are still some people who say, why wouldn't we do
all of this on US soil? For as long as possible.
Then there's the China question, and that has been the
sort of central debate in the development of US global
export controls on semi connector chips that really ramped up
during the Biden administration, and now the Trump administration has
(05:56):
taken a very different tack, where the question is not,
you know, will we ever sell chips to markets in
the Middle East or in Southeast Asia, but what is
the risk that those chips could ultimately benefit China, either
by Chinese AI companies remotely accessing data centers in third
countries to train AI models from the physical hardware being
(06:18):
actually diverted and transshipped to China, or from those countries
taking American hardware and AI know how and then making
technically separate but kind of related investments in Chinese tech.
And so that's the rub, and the question is what
conditions is the US going to put on these deals
to ensure that you're safeguarding American technology and know how
(06:41):
and ensuring that in diffusing the technology around the world,
you're not inadvertently and benefiting a US adversary.
Speaker 2 (06:47):
World quick specific question, are there stipulations in the deal
that setting aside the implementation of it that are designed
to create safeguards or checkpoints such that the chips that
go to the golf are then transship to Chinese data centers,
Like do they have some ostensible mechanism for avoiding that?
Speaker 4 (07:06):
Yes, the US government has a small handful of export
control officers that operate globally, and by small I mean
like very low double digits that can go in and
like check physical facilities and see, okay, are the chips
still here? They talk to the data center operators and such.
The Trump administration has asked for more funding for the
(07:27):
agency that executes those controls, and there are high level
provisions in this bilateral arrangement between the US and the
UAE about preventing diversion to China. There are also high
level provisions about restricting China's remote access to facilities and
around inbound and outbound investment from China reciprocated with the
(07:48):
Gulf nations in question. Those are high level arrangements, and
the big company deals were announced when the security arrangements
remained at a fairly high level. And so there are
some folks in Washington who are concerned that the US
didn't hammer out all of the specifics dot every I,
cross every tea before announcing these deals and topping them
(08:11):
as major wins for American economic competitiveness and national security.
Speaker 2 (08:30):
So it seems like when we talk about the US
China AI rivalry, there's two potential dimensions, which is one
is there's this sort of anxiety like China gonna achieve
AGI first or is it gonna raise ahead of you know,
are their labs gonna be better than our labs, et cetera.
But then there seems to be this other one and actually,
like David Sachs, the All In podcast host but also
(08:51):
Trump's as are, seems to have a slightly different spin
on the question, which is just like, no, like we
really should just want as much US tech currently diffused
out in the world, that the dominance of the Kuda
ecosystem and Nvidia, et cetera is what we should really
care about, and that we're cutting our shooting ourselves at
the foot if we get too nervous about where our
chips are.
Speaker 3 (09:11):
Going in the world.
Speaker 4 (09:12):
David Sachs basically defined winning in the AI race as
capturing the vast majority of the global market for AI hardware,
and of course, the US ability to do. That is
a function of Vidia's dominance and US restrictions that have
been levied for years on China's semiconductor industry and its
ability to build a chip making prowess. So we've seen
actually a couple of estimates recently of what china se
(09:35):
conductor landscape looks like. You had Commerce Secretary Howard Lutnik
tell the Senate Appropriations Committee last week. He was asked
by a senators and oh, I think I've read that
China can make less than four hundred thousand advanced chips
a year, And Lutinix said, actually, no, they say they're
making them, and they're not. It's probably closer to two
hundred thousand. He didn't define what chips he was talking about.
(09:57):
Advanced chips could include mobile processors, for example, in addition
to AI chips, But for the sake of discussion, let's
say it's waways ascend AI chips, which are the best
domestic model in China. That number is really far short
of China's domestic demand, which the market research firm tech
Insights put at around one point five million AI accelerators
(10:20):
in twenty twenty four. Compared to what Lennink says is
two hundred thousand that China can domestically produce. There's another
estimate out there from the research firm Semi Analysis that
says China can produce around three hundred and eighty thousand
of its most advanced AI chips domestically and that will
reach multiple millions next year. Now, in addition to China's
(10:41):
domestic chip making capacity, there is also a significant infusion
that Huawei got from ordering a reported two point nine
million chips from TSMC, the global leader in actual chip manufacturing,
through an intermediary that has since been sanctioned by the
US government. And that's not an insignificant one time infusion
(11:05):
into China's stockpile. Now, the Biden administration released a rule
in their final week in office increasing due diligence requirements
for foundries, which are the companies that actually make chips
designed by others. And TSMC hasn't commented publicly on the
two point nine million number, but they said that they're
cooperating with a US investigation into the matter. So the
(11:25):
question for the US is not will China ever develop
good chips that can compete on a global scale. I
don't think that policymakers from the Biden administration to the
Trump administration have any reservations really about thinking, of course,
China's a serious competitive technological player. It's about when that
(11:47):
could happen, and when that could happen in a way
that would allow China to export in significant quantities to
foreign markets. So there's this really interesting example that happened
a couple of weeks ago where the Deputy Communications Minister
in Malaysia, in a speech that was picked up in
local media, said we have this new national AI system
(12:08):
that's powered by three thousand Huawei Ascent chips. Now, three
thousand chips is like a super small number of chips.
It's a fraction of what's needed for a mid sized
data center. But it was the first known example, or
at least one of the first note examples, at least
the first that I attract of an international deployment from China.
(12:29):
And so there was this big reaction in DC of like,
oh my god, it's happening. David Zachs tweeted, as I've
been warning, the full Chinese stack is here. We rescinded
the Biden diffusion rule just in time. That's a reference
to global semiconductor controls that the Biden administration imposed on
pretty much the entire world in their last week in office,
(12:49):
and David sax finished, the American AI stack needs to
be unleashed to compete. So I was super interested in this.
My colleagues in kual Umpur were super interested in this,
and so we called the Malaysian Communications Ministry. We said, hey,
can we get some more information about this project and
they said, oh, we're retracting the Wawei part of that speech.
Now important in context for this is that it came
(13:11):
a couple of days after the US Commerce Department had
issued guidance saying that it considers the use of Huawei's
ascendships anywhere in the world to be a violation of
US export controls. Now they've since removed that anywhere in
the world. Line, it's like this whole thing and the
ongoing units China US China trade negotiations. But that's an
important piece of understanding Malaysia's reaction here. So we published
(13:37):
a story that was basically like Malaysia's distancing itself from
this Huawei project. And the response that David Sachs had
to that after on Twitter was that's not a lot
of chips, but it does show China's intent to export
the Huawei plus Deep Seek stack. Deep Seek was also
part of Malaysia's approach, and it's a harbinger of what's
(13:58):
to come.
Speaker 3 (14:00):
Now.
Speaker 4 (14:00):
Malaysia is like, we're going to assert our sovereignty. We
get to decide which technology we want to use. But
it was really interesting to watch the reaction to this
play out in DC because, on the one hand, some
people treated it as evidence that the threat they've been
warning about is here and the US therefore needs to
(14:22):
be super aggressive about exporting Nvidia chips as quickly and
as widely and in as large quantities as possible. And
there are some people who read it as this is propaganda,
This is Huawei trying to demonstrate that it can export
internationally when really it can't even meet domestic demand. Now,
last thing I'll say here before I turn it back
over to you guys, is there was a really interesting
(14:46):
interview with Wawai's founder in People's Daily, the official newspaper
of the Chinese Communist Party, just I guess today in
Hong Kong yesterday in DC, depending time zone you're in,
where he said Huawei's actually still behind the US. US
overestimating our capabilities, but we're making progress now. It will
always still constrained, in large part due to US restrictions
(15:08):
on semiconductor manufacturing equipment.
Speaker 3 (15:11):
Since you brought up the global export controls on chips,
it seems to me like there's a fundamental tension here
where if you want to export US technology, you kind
of have to accept that some of the risk is
that it's going to leak into parts of the world
where you know, you might not want it to be,
so China, And when I imagine, you know, if the
(15:33):
US says it wants to release its tech stack globally, sorry,
I have an image in my mind, it's like, release
the hounds, release the tech stacks. If they want to
do that, it seems like I guess they either have
to fine tune the export controls or harden them up.
And I'm not entirely sure how they would do that,
given that we already had the experience of the Biden
(15:56):
administration's export controls, Like what levers can they actually pull
here to ensure that American technology goes global, But at
the same time, it doesn't go global in a way
that we don't want it to.
Speaker 4 (16:10):
It's a really challenging question. I mean, that's why you
saw the Biden administration essentially pause large scale Nvidia and
am Dai chip shipments for the better part of a year.
You know, the US restrictions the like. Really US restrictions
on AI chips started in October twenty twenty two, a
date that anybody in the Chinese tech industry would remember
when the US said all right, no more in Vidia
(16:31):
exports of China without a license at the highest end
of video technology. That's a month before chad GPT launched.
And if you think about what's happened in the you know,
nearly three years since then, there's this massive interest from
countries all over the world in developing what's you know,
people in the industry called sovereign AI, basically the idea
(16:52):
that as a nation, you have the hardware and software
infrastructure to develop, train, deploy your own AIM models. And
the Gulf is a really interesting first case study of
a really organized ambition to make that happen. But there
are countries all over the world that want that. In
video CEO Trens and Wong has been talking about that
for a very long time. Mean, Sam Altman has been
(17:14):
going to the Middle East looking for funding for US
infrastructure projects, of course, setting up what became this five
gigawat data center in Abu Dhabi, and the Biden administration,
after they restricted shipments to China in twenty twenty two,
followed up a year later expanding the restrictions not just
on the technology that can go to China, but also
(17:37):
the global footprint of the controls, and they added around
forty countries on the basis of the risk of diversion
to China. And then about a year after that, in
their very final week in office, they released this global
framework called the AI Diffusion Rule that basically split the
world into three categories of chip access. In Tier one,
(18:00):
you had the US and very close partner countries that
the Biden administration viewed to have similar export control regimes
and capacities to those offered by Washington. In Tier three
the bottom, where you have your China, Russia or on
North Korea US adversaries that basically just continue to not
have effectively access to advanced American tech. And then you
(18:21):
had everybody in the middle, which was most of the world,
everybody from treaty allies like Poland to Singapore to Yemen,
and the Biden administration's approach was, we're going to cap
the number of chips that can go to these countries.
But a company can apply for a license called a
(18:44):
validated end user designation, which is fancy Washington speak for basically,
I'll submit to you the security requirements that I'm going
to apply to a data center that I'm going to
build an X, Y, and Z place, and in exchange,
you were going to can make it easier for me
to get licenses to ship to that place in the future.
And because I've made all of those security promises, my
(19:07):
projects are going to be exempted from the national cap
in this place. And the Biden administration's goal was to
set that cap at a level that incentivized companies like
Microsoft or like Oracle to agree to a bunch of
security conditions, apply for licenses, and then sure, go ahead
(19:27):
build your data centers in the places that you want to,
as long as you keep most of your compute capacity
in the US and friendly countries, and no more than
seven percent of your computing capacity in any one country
that is in another tier. So this immensely complicated regulation
got a lot of blowback from all over the tech industry.
Although a couple of supporters of some of the provisions
(19:48):
in Nvidia will surprise nobody was one of the strongest opponents.
Oracle was also a huge opponent because that cap on
putting more than seven percent of your capacity in any
one country was a big problem for Oracle's massive plan
data center Expanision in Malaysia. So you had this big
jockeying for position among tech companies and also among countries,
(20:09):
some of whom were like, we cannot execute the plans
that we want to within the constraints of this cap.
Others of whom didn't necessarily have super ambitious data center plans,
but we're kind of like, what the heck, Like, why
are we being put in the same category as this
country or this country or this country. I mean, you
had a couple of European Union nations that were put
in Tier two, but most of them were put in
(20:31):
the most preferential tier, and so the EU is like, Okay,
we're trying to disrupt free trade within the block. And
that debate sort of carried out all throughout the first
couple of months of the Trump administration. Before the Trump
officials decided all right, we're acting that rule and we're
going to work on a replacement. And the deals that
we saw happen in the UAE and Saudi Arabia are
(20:54):
the best indication that we have so far of what
that replacement could look like. And we're shifting from this,
you know, big global regulation to a lot of country
specific deals that will probably have some commonalities but will
really be negotiated on a bilateral basis.
Speaker 2 (21:11):
The absolutely fascinating topic, and you explain it very clearly,
Mackenzie Hawkins. Thank you so much for coming back on
odd Lots. Thank you so much for staying up late
for the recording, and talk to you soon hopefully.
Speaker 4 (21:23):
Thanks much for having me.
Speaker 2 (21:39):
I find this story to be extremely fascinating for multiple reasons.
I guess one is like, you know, when we think
about like export controls, or I think about what you sell,
you know, you have to think like traditionally like defense tech, right,
and so it's like, okay, you have to make a
deal at the high level before one country can buy
jets or whatever from you. But this is AI is
so weird because it's also just this commercial tech that
(22:02):
companies are going to use in random chatbots and software
as a service applications, et cetera. And yet we still
sort of treat it as potentially being defense like in
its implication.
Speaker 3 (22:15):
Yeah, it's kind of funny. I'm thinking now, imagine like
reading one of the stories about AI and China US
competition and replacing like chatbots AI with ballistic missiles, and
like what that language would actually read like. But okay,
two things here. So it is true people talk about
the race for AI supremacy in these sort of militaristic terms,
(22:37):
which obviously raises the question of what exactly success looks
like in this sphere. And you brought up the point that,
you know, some people seem to think it's achieving AGI
for some people like Lutnik, and some people like sex
seem to think that it's basically taking market share or
having dominant market share. And I don't know, Like, it's
(23:00):
strikes me that there's there's probably not going to be
a moment when the US can declare victory or China
can declare victory, but we should at least have some
sort of like agreed idea of where we're heading here.
And then the second thing I was thinking about is
again because all of this is constantly couched in militaristic
language and basically competition between the world's two biggest economies.
(23:24):
You can think of data centers as critical infrastructure, right, Yeah,
And so I guess the question is should we be
building more critical infrastructure in a place like the UAE.
On the one hand, you know, the US and the
UAE have a strong partnership. There's a US military presence
in the UAE already, US border control at the airport,
(23:44):
which is very, very convenient. But you know, we also
get oil from the Middle East, and I kind of
wonder if we want to like double down on our
infrastructure related dependence on that particular region.
Speaker 2 (23:59):
So on the one hand, the day Sacks view is like,
this is a business right and the goal is to
you know, have greater market share, et cetera. But part
of the reason anyone is even talking about this and
such and where every country needs to have their own
compute is because there is this perception that it's existential technology.
So the two sides like they sort of play off
(24:20):
of each other this. You know, there's always this sort
of like Baptist and bootleggers sort of way. I think
about this, where both sides of these questions end up
boosting each other because David's view is like, Oh, all
of this, all of this worrying is wrong, and it's
just it's a business and we want to have more
market share, and all of the he's very critical of,
(24:41):
like the AI doomers who think it's going to destroy
all the jobs, because he claims, oh, this is a
trojan horse for more regulation of the tech, et cetera.
And yet if it weren't for the AI doomers, if
it weren't for all the people thinking this is existential technology,
the wouldn't be that much demand to just absolutely have
this many chips within your bod, etc. So in all
(25:01):
of these stories with AI, it's like there's different perspectives,
and even people on opposite sides of the debate end
up like bolstering the others beating each other in anyway,
I think these are interesting stories to watch. I had
missed that whole thing with Huawei ostensibly and then Malaysia
backtracking about its involvement, but certainly their AI chips being
(25:25):
used globally would be a very interesting development to watch.
Speaker 3 (25:28):
Yeah, lots more to talk about on this topic, I'm
sure shall we leave it there for now?
Speaker 2 (25:32):
Let's leave it there.
Speaker 3 (25:33):
This has been another episode of the Odd Thoughts podcast.
I'm Tracy Alloway. You can follow me at Tracy Alloway.
Speaker 2 (25:38):
And I'm Joe Wisenthal. You can follow me at the Stalwart.
Follow our guest Mackenzie Hawkins, She's at mac Hawk. Follow
our producers Kerman Rodriguez at Kerman armand dash Ol Bennett
at dashbod and kil Brooks at Cale Brooks. From our
odd Lots content, go to Bloomberg dot com slash odd Lots,
where have a daily newsletter and all of our episodes,
and you can shut about all of these topics twenty
(25:59):
four seven in our discord Discord dot gg slash.
Speaker 3 (26:03):
Odd Lots And if you enjoy odd Lots, if you
like it when we talk about AI campuses in appudapi,
then please leave us a positive review on your favorite
podcast platform. And remember, if you are a Bloomberg subscriber,
you can listen to all of our episodes absolutely ad free.
All you need to do is find the Bloomberg channel
on Apple Podcasts and follow the instructions there. Thanks for
(26:24):
listening
Bloomberg Audio Studios, Podcasts, radio News.
Speaker 2 (00:18):
Hello and welcome to another episode of the Odd Lots podcast.
Speaker 3 (00:21):
I'm Joe Wisenthal and I'm Tracy Alloway.
Speaker 2 (00:24):
It's a few weeks ago, but last month Trump was
in the was in Saudi Arabia, he was in the
Middle East, and he sold a bunch of chips.
Speaker 3 (00:30):
Yeah, I think the chip's announcement was kind of overshadowed
by some of the pomp and ceremony around that visit.
Did you see the famous hair flipping No, oh, you
didn't see that, So you need to look this up.
So when Trump arrived in the UAE, part of his
big you know, the greeting and the red carpet getting
rolled out for him was a bunch of emmiati women
(00:52):
doing a hair flipping dance. I gotta say, I wish
I could flip my hair like that, but sadly I
could not.
Speaker 2 (00:59):
Can I say, like Trump seems very at home in
the Gulf States, you know what I'm saying, Like it
feels like, setting aside many other things that there is
like a similar sense of esthetics.
Speaker 3 (01:11):
Grant I was about to say that he's like, I think.
Speaker 2 (01:14):
He's very comfortable in those countries.
Speaker 3 (01:16):
I was about to say, his taste in interior decorating
is exactly the same as what you see in palaces
and hotels in Saudi Arabia and places like that.
Speaker 2 (01:25):
Absolutely, No, I think just the whole vibe is like, oh,
his on home territory. But I think the AI stuff
is really interesting. Obviously, for a lot of reasons, the
US wants to sell more things. One of the things
that we can sell in various flavors is AI technology.
But there's this weird like I don't know what the
word cross currents or whatever, which is that we want
to sell more stuff, and yet we also don't want
(01:48):
our stuff to be in the hands of perceived geopolitical rivals,
and so we cut off our own sales. China is
obviously the most salient example here, but it sort of
makes this interesting tension when we think of the role
of AI sales tech sales in sort of international relations
and diplomacy.
Speaker 4 (02:04):
Yeah.
Speaker 3 (02:04):
Well, I also find this interesting from the UAE slash
Abu Dhabi side of things, right, because the UAE, it's
strategically aligned with the US in many ways, but it's
very commercially aligned with China, and you know, you have
a lot of Chinese commerce coming through the ports and
things like that, so I'm very curious what they see
(02:26):
in this steal and how they're navigating those two sides
as well.
Speaker 1 (02:29):
Well.
Speaker 2 (02:29):
I'm very excited we have the perfect guest, someone one
of our colleagues we've had on the podcast before. We're
going to be speaking with Mackenzie Hawkins, tech and geopolitics
reporter Bloomberg News. She's now based in Hong Kong, sadly
for us, but thankfully she's you know, staying up late
for this recording. Mackenzie, thank you so much for coming
back on odd locks.
Speaker 4 (02:49):
Thanks so much for having me.
Speaker 2 (02:50):
First of all, why don't you just give us the
specific details, as the salient details as you understand them
in terms of this big, big AI deal that Trump
announced when he did his golf tour.
Speaker 4 (03:06):
So in the four day span or so that Trump
was in the Middle East, you saw American tech companies
and companies in golf nations announce hundreds of billions of
dollars worth of AI deals. Specifically, many of them focused
on AI infrastructure, building out the data centers that house
AI chips that are used to train and deploy large
language models, and that involved one of the biggest data
(03:30):
centers potentially in the world, a project anchored by the
Abu Dhabi AI firm G forty two, with key participation
from Oracle, from Open AI, from Nvidia, huge volumes of
Nvidia ship sales going to both the UAE and Saudi Arabia,
as well as sales from Nvidia's rival Advanced micro Devices,
and all told, we sort of ended the week, and
(03:53):
the deals announcements actually continued into the following week. We
ended this trip going from a world where the two
main players with significant AI data center capacity the US,
of course, and then China also has major ambitions and still,
even if all of these deals fall through as expected,
which is an if, the US will maintain the vast
(04:15):
majority of the world's computational power, but we have two
new serious players in the UAE and in Saudi Arabia.
Speaker 3 (04:23):
One thing I know from living in Apudapi for a
couple of years is that Apudapi loves building a giant
campus with some sort of foreign partner like that is
always happening and seems to be happening more and more.
But okay, so from the US side. You know, the
US wants to be a powerful player in AI and
everything that comes with it, chips, data centers, et cetera.
(04:44):
And they want to sell that technology to other countries.
What's the issue with the deal? Why is it controversial?
Speaker 4 (04:51):
So there are a couple of sort of categories of
disagreement in Washington, if you The first is a basic
economic competitive this argument. You have people, especially on the
Democratic side of the aisle, saying, you know, why are
we building a data center in Abu Dhabi when we
could be doing that in Ohio or in Pennsylvania. The
response from people who are in favor of these deals is,
(05:12):
we are building data centers in the US. Trump is
trying to cut red tape to facilitate data center deals
across the country, and the UAE accord, in particular, because
there's now a bilateral arrangement between American and AMRADI officials,
includes this reciprocal investment portion where you know, for the
amount that US companies are investing in the Gulf nation,
we'll get a reciprocal amount on American soil. But there
(05:35):
are still some people who say, why wouldn't we do
all of this on US soil? For as long as possible.
Then there's the China question, and that has been the
sort of central debate in the development of US global
export controls on semi connector chips that really ramped up
during the Biden administration, and now the Trump administration has
(05:56):
taken a very different tack, where the question is not,
you know, will we ever sell chips to markets in
the Middle East or in Southeast Asia, but what is
the risk that those chips could ultimately benefit China, either
by Chinese AI companies remotely accessing data centers in third
countries to train AI models from the physical hardware being
(06:18):
actually diverted and transshipped to China, or from those countries
taking American hardware and AI know how and then making
technically separate but kind of related investments in Chinese tech.
And so that's the rub, and the question is what
conditions is the US going to put on these deals
to ensure that you're safeguarding American technology and know how
(06:41):
and ensuring that in diffusing the technology around the world,
you're not inadvertently and benefiting a US adversary.
Speaker 2 (06:47):
World quick specific question, are there stipulations in the deal
that setting aside the implementation of it that are designed
to create safeguards or checkpoints such that the chips that
go to the golf are then transship to Chinese data centers,
Like do they have some ostensible mechanism for avoiding that?
Speaker 4 (07:06):
Yes, the US government has a small handful of export
control officers that operate globally, and by small I mean
like very low double digits that can go in and
like check physical facilities and see, okay, are the chips
still here? They talk to the data center operators and such.
The Trump administration has asked for more funding for the
(07:27):
agency that executes those controls, and there are high level
provisions in this bilateral arrangement between the US and the
UAE about preventing diversion to China. There are also high
level provisions about restricting China's remote access to facilities and
around inbound and outbound investment from China reciprocated with the
(07:48):
Gulf nations in question. Those are high level arrangements, and
the big company deals were announced when the security arrangements
remained at a fairly high level. And so there are
some folks in Washington who are concerned that the US
didn't hammer out all of the specifics dot every I,
cross every tea before announcing these deals and topping them
(08:11):
as major wins for American economic competitiveness and national security.
Speaker 2 (08:30):
So it seems like when we talk about the US
China AI rivalry, there's two potential dimensions, which is one
is there's this sort of anxiety like China gonna achieve
AGI first or is it gonna raise ahead of you know,
are their labs gonna be better than our labs, et cetera.
But then there seems to be this other one and actually,
like David Sachs, the All In podcast host but also
(08:51):
Trump's as are, seems to have a slightly different spin
on the question, which is just like, no, like we
really should just want as much US tech currently diffused
out in the world, that the dominance of the Kuda
ecosystem and Nvidia, et cetera is what we should really
care about, and that we're cutting our shooting ourselves at
the foot if we get too nervous about where our
chips are.
Speaker 3 (09:11):
Going in the world.
Speaker 4 (09:12):
David Sachs basically defined winning in the AI race as
capturing the vast majority of the global market for AI hardware,
and of course, the US ability to do. That is
a function of Vidia's dominance and US restrictions that have
been levied for years on China's semiconductor industry and its
ability to build a chip making prowess. So we've seen
actually a couple of estimates recently of what china se
(09:35):
conductor landscape looks like. You had Commerce Secretary Howard Lutnik
tell the Senate Appropriations Committee last week. He was asked
by a senators and oh, I think I've read that
China can make less than four hundred thousand advanced chips
a year, And Lutinix said, actually, no, they say they're
making them, and they're not. It's probably closer to two
hundred thousand. He didn't define what chips he was talking about.
(09:57):
Advanced chips could include mobile processors, for example, in addition
to AI chips, But for the sake of discussion, let's
say it's waways ascend AI chips, which are the best
domestic model in China. That number is really far short
of China's domestic demand, which the market research firm tech
Insights put at around one point five million AI accelerators
(10:20):
in twenty twenty four. Compared to what Lennink says is
two hundred thousand that China can domestically produce. There's another
estimate out there from the research firm Semi Analysis that
says China can produce around three hundred and eighty thousand
of its most advanced AI chips domestically and that will
reach multiple millions next year. Now, in addition to China's
(10:41):
domestic chip making capacity, there is also a significant infusion
that Huawei got from ordering a reported two point nine
million chips from TSMC, the global leader in actual chip manufacturing,
through an intermediary that has since been sanctioned by the
US government. And that's not an insignificant one time infusion
(11:05):
into China's stockpile. Now, the Biden administration released a rule
in their final week in office increasing due diligence requirements
for foundries, which are the companies that actually make chips
designed by others. And TSMC hasn't commented publicly on the
two point nine million number, but they said that they're
cooperating with a US investigation into the matter. So the
(11:25):
question for the US is not will China ever develop
good chips that can compete on a global scale. I
don't think that policymakers from the Biden administration to the
Trump administration have any reservations really about thinking, of course,
China's a serious competitive technological player. It's about when that
(11:47):
could happen, and when that could happen in a way
that would allow China to export in significant quantities to
foreign markets. So there's this really interesting example that happened
a couple of weeks ago where the Deputy Communications Minister
in Malaysia, in a speech that was picked up in
local media, said we have this new national AI system
(12:08):
that's powered by three thousand Huawei Ascent chips. Now, three
thousand chips is like a super small number of chips.
It's a fraction of what's needed for a mid sized
data center. But it was the first known example, or
at least one of the first note examples, at least
the first that I attract of an international deployment from China.
(12:29):
And so there was this big reaction in DC of like,
oh my god, it's happening. David Zachs tweeted, as I've
been warning, the full Chinese stack is here. We rescinded
the Biden diffusion rule just in time. That's a reference
to global semiconductor controls that the Biden administration imposed on
pretty much the entire world in their last week in office,
(12:49):
and David sax finished, the American AI stack needs to
be unleashed to compete. So I was super interested in this.
My colleagues in kual Umpur were super interested in this,
and so we called the Malaysian Communications Ministry. We said, hey,
can we get some more information about this project and
they said, oh, we're retracting the Wawei part of that speech.
Now important in context for this is that it came
(13:11):
a couple of days after the US Commerce Department had
issued guidance saying that it considers the use of Huawei's
ascendships anywhere in the world to be a violation of
US export controls. Now they've since removed that anywhere in
the world. Line, it's like this whole thing and the
ongoing units China US China trade negotiations. But that's an
important piece of understanding Malaysia's reaction here. So we published
(13:37):
a story that was basically like Malaysia's distancing itself from
this Huawei project. And the response that David Sachs had
to that after on Twitter was that's not a lot
of chips, but it does show China's intent to export
the Huawei plus Deep Seek stack. Deep Seek was also
part of Malaysia's approach, and it's a harbinger of what's
(13:58):
to come.
Speaker 3 (14:00):
Now.
Speaker 4 (14:00):
Malaysia is like, we're going to assert our sovereignty. We
get to decide which technology we want to use. But
it was really interesting to watch the reaction to this
play out in DC because, on the one hand, some
people treated it as evidence that the threat they've been
warning about is here and the US therefore needs to
(14:22):
be super aggressive about exporting Nvidia chips as quickly and
as widely and in as large quantities as possible. And
there are some people who read it as this is propaganda,
This is Huawei trying to demonstrate that it can export
internationally when really it can't even meet domestic demand. Now,
last thing I'll say here before I turn it back
over to you guys, is there was a really interesting
(14:46):
interview with Wawai's founder in People's Daily, the official newspaper
of the Chinese Communist Party, just I guess today in
Hong Kong yesterday in DC, depending time zone you're in,
where he said Huawei's actually still behind the US. US
overestimating our capabilities, but we're making progress now. It will
always still constrained, in large part due to US restrictions
(15:08):
on semiconductor manufacturing equipment.
Speaker 3 (15:11):
Since you brought up the global export controls on chips,
it seems to me like there's a fundamental tension here
where if you want to export US technology, you kind
of have to accept that some of the risk is
that it's going to leak into parts of the world
where you know, you might not want it to be,
so China, And when I imagine, you know, if the
(15:33):
US says it wants to release its tech stack globally, sorry,
I have an image in my mind, it's like, release
the hounds, release the tech stacks. If they want to
do that, it seems like I guess they either have
to fine tune the export controls or harden them up.
And I'm not entirely sure how they would do that,
given that we already had the experience of the Biden
(15:56):
administration's export controls, Like what levers can they actually pull
here to ensure that American technology goes global, But at
the same time, it doesn't go global in a way
that we don't want it to.
Speaker 4 (16:10):
It's a really challenging question. I mean, that's why you
saw the Biden administration essentially pause large scale Nvidia and
am Dai chip shipments for the better part of a year.
You know, the US restrictions the like. Really US restrictions
on AI chips started in October twenty twenty two, a
date that anybody in the Chinese tech industry would remember
when the US said all right, no more in Vidia
(16:31):
exports of China without a license at the highest end
of video technology. That's a month before chad GPT launched.
And if you think about what's happened in the you know,
nearly three years since then, there's this massive interest from
countries all over the world in developing what's you know,
people in the industry called sovereign AI, basically the idea
(16:52):
that as a nation, you have the hardware and software
infrastructure to develop, train, deploy your own AIM models. And
the Gulf is a really interesting first case study of
a really organized ambition to make that happen. But there
are countries all over the world that want that. In
video CEO Trens and Wong has been talking about that
for a very long time. Mean, Sam Altman has been
(17:14):
going to the Middle East looking for funding for US
infrastructure projects, of course, setting up what became this five
gigawat data center in Abu Dhabi, and the Biden administration,
after they restricted shipments to China in twenty twenty two,
followed up a year later expanding the restrictions not just
on the technology that can go to China, but also
(17:37):
the global footprint of the controls, and they added around
forty countries on the basis of the risk of diversion
to China. And then about a year after that, in
their very final week in office, they released this global
framework called the AI Diffusion Rule that basically split the
world into three categories of chip access. In Tier one,
(18:00):
you had the US and very close partner countries that
the Biden administration viewed to have similar export control regimes
and capacities to those offered by Washington. In Tier three
the bottom, where you have your China, Russia or on
North Korea US adversaries that basically just continue to not
have effectively access to advanced American tech. And then you
(18:21):
had everybody in the middle, which was most of the world,
everybody from treaty allies like Poland to Singapore to Yemen,
and the Biden administration's approach was, we're going to cap
the number of chips that can go to these countries.
But a company can apply for a license called a
(18:44):
validated end user designation, which is fancy Washington speak for basically,
I'll submit to you the security requirements that I'm going
to apply to a data center that I'm going to
build an X, Y, and Z place, and in exchange,
you were going to can make it easier for me
to get licenses to ship to that place in the future.
And because I've made all of those security promises, my
(19:07):
projects are going to be exempted from the national cap
in this place. And the Biden administration's goal was to
set that cap at a level that incentivized companies like
Microsoft or like Oracle to agree to a bunch of
security conditions, apply for licenses, and then sure, go ahead
(19:27):
build your data centers in the places that you want to,
as long as you keep most of your compute capacity
in the US and friendly countries, and no more than
seven percent of your computing capacity in any one country
that is in another tier. So this immensely complicated regulation
got a lot of blowback from all over the tech industry.
Although a couple of supporters of some of the provisions
(19:48):
in Nvidia will surprise nobody was one of the strongest opponents.
Oracle was also a huge opponent because that cap on
putting more than seven percent of your capacity in any
one country was a big problem for Oracle's massive plan
data center Expanision in Malaysia. So you had this big
jockeying for position among tech companies and also among countries,
(20:09):
some of whom were like, we cannot execute the plans
that we want to within the constraints of this cap.
Others of whom didn't necessarily have super ambitious data center plans,
but we're kind of like, what the heck, Like, why
are we being put in the same category as this
country or this country or this country. I mean, you
had a couple of European Union nations that were put
in Tier two, but most of them were put in
(20:31):
the most preferential tier, and so the EU is like, Okay,
we're trying to disrupt free trade within the block. And
that debate sort of carried out all throughout the first
couple of months of the Trump administration. Before the Trump
officials decided all right, we're acting that rule and we're
going to work on a replacement. And the deals that
we saw happen in the UAE and Saudi Arabia are
(20:54):
the best indication that we have so far of what
that replacement could look like. And we're shifting from this,
you know, big global regulation to a lot of country
specific deals that will probably have some commonalities but will
really be negotiated on a bilateral basis.
Speaker 2 (21:11):
The absolutely fascinating topic, and you explain it very clearly,
Mackenzie Hawkins. Thank you so much for coming back on
odd Lots. Thank you so much for staying up late
for the recording, and talk to you soon hopefully.
Speaker 4 (21:23):
Thanks much for having me.
Speaker 2 (21:39):
I find this story to be extremely fascinating for multiple reasons.
I guess one is like, you know, when we think
about like export controls, or I think about what you sell,
you know, you have to think like traditionally like defense tech, right,
and so it's like, okay, you have to make a
deal at the high level before one country can buy
jets or whatever from you. But this is AI is
so weird because it's also just this commercial tech that
(22:02):
companies are going to use in random chatbots and software
as a service applications, et cetera. And yet we still
sort of treat it as potentially being defense like in
its implication.
Speaker 3 (22:15):
Yeah, it's kind of funny. I'm thinking now, imagine like
reading one of the stories about AI and China US
competition and replacing like chatbots AI with ballistic missiles, and
like what that language would actually read like. But okay,
two things here. So it is true people talk about
the race for AI supremacy in these sort of militaristic terms,
(22:37):
which obviously raises the question of what exactly success looks
like in this sphere. And you brought up the point that,
you know, some people seem to think it's achieving AGI
for some people like Lutnik, and some people like sex
seem to think that it's basically taking market share or
having dominant market share. And I don't know, Like, it's
(23:00):
strikes me that there's there's probably not going to be
a moment when the US can declare victory or China
can declare victory, but we should at least have some
sort of like agreed idea of where we're heading here.
And then the second thing I was thinking about is
again because all of this is constantly couched in militaristic
language and basically competition between the world's two biggest economies.
(23:24):
You can think of data centers as critical infrastructure, right, Yeah,
And so I guess the question is should we be
building more critical infrastructure in a place like the UAE.
On the one hand, you know, the US and the
UAE have a strong partnership. There's a US military presence
in the UAE already, US border control at the airport,
(23:44):
which is very, very convenient. But you know, we also
get oil from the Middle East, and I kind of
wonder if we want to like double down on our
infrastructure related dependence on that particular region.
Speaker 2 (23:59):
So on the one hand, the day Sacks view is like,
this is a business right and the goal is to
you know, have greater market share, et cetera. But part
of the reason anyone is even talking about this and
such and where every country needs to have their own
compute is because there is this perception that it's existential technology.
So the two sides like they sort of play off
(24:20):
of each other this. You know, there's always this sort
of like Baptist and bootleggers sort of way. I think
about this, where both sides of these questions end up
boosting each other because David's view is like, Oh, all
of this, all of this worrying is wrong, and it's
just it's a business and we want to have more
market share, and all of the he's very critical of,
(24:41):
like the AI doomers who think it's going to destroy
all the jobs, because he claims, oh, this is a
trojan horse for more regulation of the tech, et cetera.
And yet if it weren't for the AI doomers, if
it weren't for all the people thinking this is existential technology,
the wouldn't be that much demand to just absolutely have
this many chips within your bod, etc. So in all
(25:01):
of these stories with AI, it's like there's different perspectives,
and even people on opposite sides of the debate end
up like bolstering the others beating each other in anyway,
I think these are interesting stories to watch. I had
missed that whole thing with Huawei ostensibly and then Malaysia
backtracking about its involvement, but certainly their AI chips being
(25:25):
used globally would be a very interesting development to watch.
Speaker 3 (25:28):
Yeah, lots more to talk about on this topic, I'm
sure shall we leave it there for now?
Speaker 2 (25:32):
Let's leave it there.
Speaker 3 (25:33):
This has been another episode of the Odd Thoughts podcast.
I'm Tracy Alloway. You can follow me at Tracy Alloway.
Speaker 2 (25:38):
And I'm Joe Wisenthal. You can follow me at the Stalwart.
Follow our guest Mackenzie Hawkins, She's at mac Hawk. Follow
our producers Kerman Rodriguez at Kerman armand dash Ol Bennett
at dashbod and kil Brooks at Cale Brooks. From our
odd Lots content, go to Bloomberg dot com slash odd Lots,
where have a daily newsletter and all of our episodes,
and you can shut about all of these topics twenty
(25:59):
four seven in our discord Discord dot gg slash.
Speaker 3 (26:03):
Odd Lots And if you enjoy odd Lots, if you
like it when we talk about AI campuses in appudapi,
then please leave us a positive review on your favorite
podcast platform. And remember, if you are a Bloomberg subscriber,
you can listen to all of our episodes absolutely ad free.
All you need to do is find the Bloomberg channel
on Apple Podcasts and follow the instructions there. Thanks for
(26:24):
listening
Hosts And Creators
Joe Weisenthal
Tracy Alloway
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy And Charlamagne Tha God!
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.