August 13, 2025 • 46 mins
Advanced semiconductors, the tiny components crucial for everything from smartphones to national defense, are at the heart of US-China geopolitical rivalry. It's no wonder, then, that the US has increasingly restricted China's access to these chips, while Beijing fights back with its own trade barriers.
So where are we in the great chip war? Why are they so important and difficult to make? And what are the prospects for new alliances as chip-hungry countries navigate the new AI world? Chris Miller, author of Chip War and professor of international history at Tufts University, sits down with John and Katia for a crash course on this technology and what lies ahead.
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 (00:01):
Semiconductors are at the heart of America's geopolitical rivalry with China.
They're vital to products we use every day, including smartphones
and computers, and for advanced technologies like AI, electric vehicles
and defense systems. China's made in twenty twenty five policy
a decade ago called for aggressive manufacturing and self sufficiency
(00:23):
in this space.
Speaker 2 (00:25):
That's been worrying Washington ever since, with both the Biden
and Trump administrations adding tougher rules like cutting access to
American technology and certain semiconductors, and it's been met with
limited success. Chinese firms are still finding ways of getting
the hands on US and Taiwanese advanced semiconductors, and the
Chinese government continues to double down on its own industry.
Speaker 1 (00:49):
So where are we in a great chip war? Who's winning?
If anyone? Could we see more thawing in the relationship
as China and the US try to negotiate a trade deal,
and how can regions like Taiwan, South Korea and Japan
benefit from this rivalry. You're listening to Asia Centric from
Bloomberg Intelligence. I'm Katydmitreva in Hong Kong.
Speaker 2 (01:11):
I'm John Lee also in Hong Kong.
Speaker 1 (01:13):
And today we have Chris Miller, who wrote the book
on Justice. He's a professor at Toft University and the
author of Chip War, The Fight for the World's Most
Critical Technology, often seen as required reading for understanding this space.
He joins us in Boston. Thank you for joining us, Chris,
Thank you for having me.
Speaker 2 (01:34):
Chris in your own words, why are semiconductors so important
to the great powers?
Speaker 3 (01:40):
Well, I think there are two reasons why. The first
is that chips are just important for the global economy
writ large. If you look at the world's biggest companies
by market capitalization, they're all either designers of chips, producers
of chips, or companies that really couldn't function without chips.
Speaker 4 (02:00):
Look at global trade.
Speaker 3 (02:01):
The largest single flow in all of global trade is
the flow of chips into China each year, from Taiwan,
from Korea, from Southeast Asia. And so you can't understand
modern economies without chips. That's one reason. The second is
that when you look forward towards the application of artificial
intelligence to every aspect of the economy, what you find
(02:22):
here too, chips are central the advances that we've seen
in AI over the past couple of years have been
driven by better and better semiconductors, which are both used
to train these ultra capable AI systems but also in
deploying them at scale. And so, just like all of
the world's governments, tech CEOs too are betting that if
(02:42):
you want to be able to build and deploy the
best AI systems, you've got to be able to access
to large numbers of high quality AI chips. And so
these two factors, the macroeconomic centrality and the importance to AI,
have put chips not just on the radar screens of
Silicon Valley leaders, but also of the world's top political officials.
Speaker 1 (03:03):
And just sort of a historical point for those who
haven't read the book yet, when did this all begin?
In terms of this importance of chips, Well.
Speaker 3 (03:13):
In some ways it dates back to the origins of
the chip industry. The first ships were invented in the
late nineteen fifties. The aim was to miniatureize computing powers.
You could fit a computer in the nose cone of
a missile and guide it more accurately towards its target,
and this was critical for the Cold War competition between
the United States and the Soviet Union. But it really
(03:35):
kind of kicked up a notch about ten years ago
when the United States was the dominant player in the
chip supply chain, not the dominant manufacturer of chips, but
the dominant producer of many of the key inputs into chips,
the ip the design know how, many of the key tools,
and China realized that it's economy was becoming more and
more dependent on ships that at the time it was
(03:58):
almost completely an importer of. It produce many at all
domestically in the chips that it did produce were all
low end, and so in twenty fourteen, China launched a
major industrial policy effort trying to develop more self sufficiency
in semiconductors. It set up a very large central government
investment fund, and each of the provinces set up their
(04:20):
own investment funds to try to build a self sufficient industry.
And this was seen not just by the US but
also by other key players Taiwan, Korea, Japan is a
economic and potentially down the road, a security threat.
Speaker 2 (04:34):
And where is China now in its aims to become
self sufficient? How would you grade the efforts so far?
Speaker 3 (04:42):
Well, the chip industry is large and complex, and you
get a different answer depending on which corner of the
hip industry.
Speaker 4 (04:48):
You look at. So there's different types of chips.
Speaker 3 (04:51):
There's processor chips, there's memory ships, and there's also different
segments of the supply chain, the design process, the manufacturing itself,
the production of the tool rules that are used to
make chips, and you need to kind of look at
each of these segments individually to really understand how China
stacks up against other competitors. At a high level, I
would say this, On the one hand, China's import of
(05:14):
chips from the rest of the world, which means largely
from Taiwan and Korea and their assembly facilities in Southeast Asia,
generally hits a new high each year it's grown for
most of the last decades. China is importing larger numbers
at least larger dollar values, of chips from abroad. But
China is also producing more at home as well, and
so when it comes to lower range and mid range ships,
(05:36):
China's building a lot of self sufficiency in the types
of ships that go into cars and to refrigerators and
industrial equipment, but the high end ships are getting even
more important. Both the chips that go in phones and PCs,
but also critically the processors and the memory ships that
make AI possible in data centers, and for these cutting
edge ships, China is still mostly importing the chips that
(06:00):
its tech firms need to deploy advanced computing at scale.
Speaker 1 (06:05):
So how far behind is China behind global leaders in
terms of being self sufficient?
Speaker 3 (06:13):
Well, it really depends again on what exactly one wants
to measure. If you ask how far behind is China
when it comes to designing smartphone processors, I would say
it's hardly behind at all. The smartphone processors that Huawei's
high silicon arm designed are by a lot of metrics,
almost as good or just as good as what Apple
(06:33):
or Qualcom would design, And for every different category of
chip you get a different rough metric. But there's a
couple places where China remains meaningfully behind, and it's been
struggling to replicate the capabilities of industry leaders. And I'll
flag two of them which are the most critical. The
first is in the production of the machines that make chips.
(06:56):
If you go inside a chip making facility in China today,
to the complex machines, especially at the high end, are
imported imported from Japan, from the Netherlands and from US firms,
and China has been trying to build out its own
suite of machines that can make chips. But this is
extraordinarily complex equipment, and it's been struggling to make a
(07:17):
lot of progress in replacing foreign chip making tools with
domestic variants. And because of that, China's also struggled to
ramp up cutting edge capacity to manufacture chips itself. And
so you can look, for example, at the leading Chinese firm, SMIC,
which is a chip maker, and compare it to the
leading Taiwanese firm, which is the world's leader, t SMC,
(07:39):
and they have similar manufacturing trajectories. And so you can
look at what SMIC does this year and look at
what TSMC did several years ago and measure the gap
between them as a way of assessing where China stacks up.
And here you find that SMICK right now in Shanghai
has got a seven nanimeter process that it has just
(08:00):
ramped at scale. It's producing ai chips, producing smartphone processors.
But TSMC and Taiwan did that six years ago. So
there's a five or six year gap between what Smith
can produce at scale and what TSMC can produce a scale.
So those are two segments where it's clear there is
still a significant gap between China's capabilities and the leading
Taiwanese US Japanese firms. But those are only two ways
(08:24):
you could measure the gap. And in other segments, especially design,
China's got very impressive capabilities, and I would argue it
has been closing the gap in recent years.
Speaker 1 (08:36):
Maybe a technical point, but you mentioned that China kind
of legs behind in the equipment that makes chips. So
could you describe just a bit why these machines are
so complex and why China is perhaps behind in it.
Because when you think of China as a manufacturer, I
mean the world's factory floor, it's really got the supply
(08:56):
chains down. So what is it about the equipment that
they're not able to do now?
Speaker 4 (09:00):
Yeah, it's a great question.
Speaker 3 (09:02):
When I first myself began studying the industry, it was
kind of a puzzle, why could Chiny to make everything
except for these tools? And the answer, it turns out,
is that these are the most complex and precise tools
humans have literally ever produced. So if you take, for example,
the most complicated and expensive tool called an extreme ultraviolet
photo lithography tool. These are tools produced by just a
(09:24):
single company in the world, ASML of the Netherlands. They
have hundreds of thousands of components, including the flattest mirrors
humans have ever made, the most powerful lasers ever deployed
in the commercial device, the capable of manufacturing at nanometer
scale that's billionths of a meter with close to perfect accuracy.
And that's just one of the tools you need to
(09:44):
make chips. There are many others that are only scarcely
less complex, and so when you start to look at
these tools, you realize that there's nothing that comes close
in terms of complexity and really any other industry. And
so it's not a surprise that China hasn't caught up
because most of these types of tools, there's just one
or two companies in the entire world that can undertake
(10:06):
this activity. And people have been ask why hasn't China
cut up and the production of extreme multi violet photo
lithography tools, Well, no one else has caught up either.
There's just one company that can do it. Others have
tried and failed because the engineering is so complicated and
the business model is so hard to make work, and
so it really is the fact that we're at the
(10:26):
frontiers of engineering the frontiers of physics that explain why
it's not easy for China to replicate this, even though
China's a course, very good at manufacturing many other types
of machinery.
Speaker 2 (10:38):
Asia Centric is produced by Bloomberg Intelligence. We're more than
five hundred experienced analysts and strategists work around the clock
to bring you timely, world class research. Our coverage spans
two hundred market industries, currencies, commodities, and industries, as well
as over two thousand equities and credits. To learn more
about Bloomberg Intelligence, visit b I go on the blue
(11:00):
the terminal. If you like what you hear, don't forget
to subscribe and chair Chris, Look, I read your book,
But for the uninitiated who didn't read your book, can
you explain why like this industry lends itself to monopolies
like you mentioned that ASML pretty much like no one
else makes these you know, lithic of methography on machines,
(11:24):
and I think you said no one else makes these
TSMC also manufactures over ninety percent of the advanced AI chips.
Why is this the case? That there are monopolies formed
in the supply chain.
Speaker 1 (11:35):
Yeah.
Speaker 3 (11:35):
I think there are two factors that lend the chip
industry to these sort of oligopolistic rin monopolistic markets. The
first is the capital expenditures are often tremendous. So for TSMC,
every new factory they build is the world's most expensive
factory because each factory is more expensive and breaks the
(11:56):
prior record. We're talking twenty or twenty five billion dollars
per FACI, and so this scope for a new startup
firm to compete is approximately zero. That has really kept
competition down because the number of players that could even
conceive of taking on TSMC is very small, and the
(12:16):
same dynamic is true in other segments of the supply chain.
Speaker 4 (12:19):
The dollar values are lower, but not much lower.
Speaker 3 (12:22):
Even if it's only three billion dollars for a new
factory that makes the ultra purified a silicon wafers, for example,
that's still a monumental capital expense for most companies, and
it deters new entrants from even trying. The second factor
is that the R and D is really extraordinarily highly specialized,
and for a lot of the process steps and making chips,
(12:44):
the only way you learn about how to undertake that
specific step is by having already been in the industry
and learning alongside your suppliers and your customers.
Speaker 4 (12:55):
And so, if you think of.
Speaker 3 (12:56):
The way the chip industry's R and D works, every year,
the entire industry works together to bring a new process
that's more capable than the prior years, often dramatically more
capable in terms of the computing power it produces. And
so you've got really tight alignment between the existing chemicals producers,
existing machinery producers, existing chip manufacturers themselves to make sure
(13:17):
that these processes arrive on time and that all the
necessary components are together. There's a lot of shared learning
inside of that small circle. But if you're an outsider
who's never been part of that learning process, there's no
way to do it on your own. You can't just
test on your own very easily whether your new chemical
will work with a existing extreme ultraviolet photogthography machine. If
(13:40):
you want to buy one of those machines yourself, it
costs you one hundred and fifty million dollars. And moreover,
companies aren't going to let you run a lot of
tests in their fab because they just spend twenty billion
dollars building a fab they.
Speaker 4 (13:50):
Need to run that at full capacity.
Speaker 3 (13:52):
So it's really hard to break in from an R
and D perspective, And when you combine that with the
capital intensivity, it means that many of the firms that
have key roles in the supply chain have had them
for the last decade or in some cases the last
three or four decades, and that has kept competition down
in the hip industry.
Speaker 1 (14:10):
So you wrote your book in twenty twenty two, that
was under the Biden administration. We're now under the Trump
administration two point zero. What are the big differences you've
seen in how both administrations have handled this issue? Are
there some similarities that come up, but also some of
the big big differences, because of course under Biden we
had the Chips IRA Act, But under both administrations it
(14:35):
seems like a bipart is an issue.
Speaker 3 (14:38):
Yeah, I think it is, and I think that the
similarities are to me more substantial than the differences. So first,
there's a focus on domestic manufacturing in the United States,
an effort to reduce reliance both on the lower end
ship produced in China, but also dependence on high end
chips produced in Taiwan the Chips Act. Although Biden claimed
credit for it, it actually emerged under the trumpetministration. That was
(15:00):
when the Congress first started discussing the bill, and people
like Secretary of State Mike Pompeo played a major role
in some of the early investments that the Taiwanese from
TSMC announced the United States. So there's been a through
line of support for manufacturing chips in the United States
that has been bipartisan.
Speaker 4 (15:17):
I think that's a point of continuity.
Speaker 3 (15:20):
I think the second point in continuity has been a desire
to stop China from accessing the most advanced ship making technologies.
The export controls on ASML tools, for example, started under
the first administration, were expanded under Biden, and we'll see
they could well be expanded further at some point under
the current TRUP administration. So that's also a through line.
(15:40):
And I think the third is concern about Chinese industrial
subsidies in the chip industry and implications for trade policy.
The Trumpdministration in round one imposed higher tariffs and Chinese
ships imported the United States. Biden expanded those tariffs, and
of course President Trump right now is considering what he's
going to do.
Speaker 4 (16:01):
On chip tariffs.
Speaker 3 (16:01):
But there's widespread expectation that there will be another round
of chip tariffs coming at some point later this year,
and so that's also a through line of concern about
Chinese subsidies. I think there are differences. Obviously, you know,
we're going to see where exactly Trump's chip tariffs land,
and they might end up being much broader on a
whole range of different countries, not just China focused.
Speaker 4 (16:21):
If so, that would be a difference.
Speaker 3 (16:22):
But I think when you zoom out, you actually end
up seeing a pheromounta continuity, despite some nuances that are different.
Speaker 1 (16:31):
Chris.
Speaker 2 (16:31):
President Trump seems to be dialing back on some of
the policies now. Under the previous administration, they banned the
export of high end in Nvidia chips into China, but
Donald Trump recently suggested that they could allow the export
of certain Age twenty chips back into the country. What
does this mean, Well, I.
Speaker 3 (16:54):
Think it's too soon to have a confident judgment. I
would say when the Biden administration left office, in Videau's
H twenty were still legal to sell China. Then in
month four of the trumpdministration, the President banned the export
of Age twenty chips, and now there's suggestions they might
be allowed again. As of today, we're still at a
tighter restrictive policy than when the Biden administration ended.
Speaker 4 (17:17):
When the Trump administration started.
Speaker 3 (17:18):
We'll see exactly how this plays out, but I think
it speaks again to a continuity, which is that both
Biden and Trump feel like they need to balance action
on chip issues with the broader.
Speaker 4 (17:32):
US China relationship.
Speaker 3 (17:33):
So into the Biden administration, it was regularly the case
that the Commerce Department would have a new regulation ready
that would limit this or that aspect of the chip trade,
and then it would wait for three months or six
months or nine months while the State Department found what
they thought was a quote unquote good window in which
to roll out a new restriction that the Chinese wouldn't like.
And cour we're seeing the trumpministration as something similar.
Speaker 4 (17:55):
Right now.
Speaker 3 (17:56):
The President wants to negotiate a trade deal with China.
Their White House is trying to get rare earth magnets
flowing again, and so there's concern in the White House
that if they were to take punitive steps on chips
or on chip making tools, that this would disrupt the
US China relationship. One can debate whether that's the right approach,
but I think it is a striking similarity to a
(18:18):
lot of the delays and the kind of half hearted
decision making that you saw in the Biden administration when
it came to whether or not to restrict certain types
of tools or chips.
Speaker 1 (18:30):
So, after the Trump administration recently allowed the export of
H twenty chips, very closely afterwards, China called in Nvidia
staff to have a discussion with them behind that is
at least partly the Trump administration suggesting to track these chips.
And this is all happening within the context of a
(18:52):
trade war and trade negotiations. So one of the questions
we've been thinking about is, you know, why are the
US and China right now fighting over what are essentially
kind of third rate chips. I mean, these aren't like
the most advanced chips out there. As John ass like,
how should we read this kind of back and forth.
Speaker 3 (19:13):
Well, I wouldn't say it's right to consider the H
twenties third rate chips. I think if you're a Chinese
tech company, these are first rate. These are the US
you can get at scale can you smuggle in small
volumes of the higher quality H one hundred yes. Can
you access small volumes of Huawei ASSN chips Yes? But
can you access either by the tens or hundreds of thousands, No,
(19:34):
both because Huawei can't produce that efficient scale and because
smuggling you can do a medium scale, but not at
the scale you need. So you know, judging by the
purchasing patterns of byte Edance and Ali Baba Intense, these
are great chips, which is why they really want to
spend hundreds of millions or billions of.
Speaker 4 (19:51):
Dollars on them.
Speaker 3 (19:52):
And I think that that is why this is such
a critical issue, both for the US and for Chinese
tech firms.
Speaker 4 (19:58):
If they really were third rate, we wouldn't be talking
about them.
Speaker 2 (20:00):
Chris like and Vidia CEO Jensen Huang was very critical
about the US ban on chips to China. He said
it hurts US businesses, and he also insinuated that it's
going to accelerate China's domestic semiconductive development. Do you agree
with him? Do you think these mergers are in some
way self defeating?
Speaker 3 (20:20):
Well, I think there's no doubt that in the short
run it reduces in video sales. That's certainly clear. So
does it hurt that business? I think the answer is yes.
One of the dilemmas that any sort of export control
regime faces is that if you restrict the sale of
a certain tool, you hurt the tool makers, but you
benefit the domestic companies that use that tool, who have
(20:42):
less foreign competition. So, for example, if you restrict the
sale of chip making tools, it's bad for the companies
that make chip making tools, but it's actually better for
chip makers in the West, including in Taiwan and South Korea,
because they don't have strong competition in China at the
cutting edge. And if you thought about the counterfactuals pose
the Trudministration had not leaned on the Netherlands in twenty
(21:03):
nineteen and said the Netherlands don't sell UV tools to China,
China would be operating today a fleet of many UV tools,
possibly the second largest behind Taiwan in the world, and
that'd be a totally different landscape for the world's chip makers.
It'd be better for sellers of chip maker tools, but
worse for chip makers. And we're in a similar dynamic
right now with chips. If you sell more chips it's
(21:24):
good for chip makers, but it's probably worse for the
companies that use chips, which are the big tech companies,
cloud computing companies.
Speaker 4 (21:31):
And AI labs. Why is it that.
Speaker 3 (21:35):
We haven't seen large deployments of AI chips buy Chinese
firms outside of China's because they can't produce enough, because
they haven't been able to access large volumes of the
high end tools that foreign firms have. And so I
think that the trade off that the US has to
navigate is, Yes, it's true that it's worse for your
(21:56):
chip makers if you don't allow sales to China, but
it's also probably better for your tech firms that use
those ships because they won't have strong Chinese competitors or
as strong Chinese competitors, and that's a trade off. I
think there's no doubt about that. I think the other
question is is US policy accelerating China's technological development?
Speaker 4 (22:15):
And here I look for empirical data points.
Speaker 3 (22:19):
We know that in twenty fourteen, China launched its first
national level Integrated Circuit Investment Fund.
Speaker 4 (22:26):
It's been re upped now twice.
Speaker 3 (22:28):
Since then at the national level, and each province has
If you ask yourself is it clear from the data,
and the data's kind of messy, but is it clear
that there's been a major increase in Chinese government subsidies
towards the ship industry after export controls? It's not really
clear because it started in twenty fourteen and it's been,
as far as we can tell, roughly steady since then.
(22:50):
If you ask, has there been a noticeable increase in
the Chinese government pushing companies to buy domestic chips after
export controls?
Speaker 4 (22:59):
I don't think it's clear.
Speaker 3 (23:00):
I think you've seen from twenty fourteen to present a
steady push to buy a local And so if you ask,
you know, what's the evidence that there's been a major
uptick in self sufficiency, either support or pressure from the
Chinese government.
Speaker 4 (23:13):
Around export controls? It's hard to find.
Speaker 3 (23:16):
Now, could you argue that domestic Chinese tech firms before
export controls would have said, of course, we're going to
test local products. They would have bought local products and
put them in the closet and instead relied on imported products.
And now if they can't import products, they've got no
choice but to take the domestic products. Seriously, I think
(23:37):
you could argue that and perhaps that's happening to some degree,
but I would say, again, the evidence for that is
not super robust. And so if you're just trying to
be empirical and say, what's been the evident changed in
Chinese policy or the approach of Chinese tech firms or
export controls, I would say it looks pretty continuous over time.
In other words, export controls didn't obviously have a major impact.
Speaker 1 (23:59):
Are the expert restrictions enough though? Are there other things
that you see the administration moving towards well?
Speaker 3 (24:06):
I think a lot depends on the broader US China relationship.
There are certainly voices in the administration that want to
do a lot more on the restrictive side. There are
also voices administration that want to do a lot more
on the promoting US technological advances side. But of course
the administration has many priorities, and it's hard to see
which is going to win out. I think in terms
of the restrictions, it's very clear that the White House
(24:29):
is going to try to balance the Hawk's desire for
restrictions with the broader US channel relationship. And so as
we approach what many people think is going to be
a summit, this fall. I think history would suggest you
shouldn't expect freemen restrictions and then run up to a
summit on the promoting technological development side. I think there
certainly are a lot of voices administration that want to
(24:49):
see more money flowing to new firms with disruptive technology.
But there's also a desire for pretty dramatic budget cuts
to science and technology in the USA, and that cuts
in the opposite direction. So how exactly these political priorities
balance out is hard to predict with any certainty.
Speaker 1 (25:07):
Just a quick follow up on the point you made
about this balance or dichotomy between US producers and exporters
of chips and tech firms, how do you see the
Trump administration to serve the next few years navigating that?
Does it mean that they need to find other ways
of supporting chip makers, for example, if they're going to
be closing the route to China.
Speaker 3 (25:30):
Well, I don't think any administration has done a great
answer to that question. I mean, China's obviously the second
large economy in the world, and you know, twenty percent
of GDP is a really large market to not be
able to serve. I think the US has tried to
argue to tech firms. Look, you can't serve this twenty percent.
You have the rest of the eighty percent you can serve.
And I think tech companies are inclined to serve that
(25:53):
other eighty percent anyway to a large degree.
Speaker 4 (25:56):
But is it right that companies.
Speaker 3 (25:58):
Would be more profitable have high revenue if they could
sell to the entire one hundred percent of GDP?
Speaker 4 (26:02):
For sure, no doubt about that, Chris.
Speaker 2 (26:05):
I'd love to talk about TSMC now. The company became
the first Asian stock to reach a market capitalization of
one trillion dollars. That was just you know recently. Can
anyone challenge TSMC's dominance in this space.
Speaker 4 (26:21):
Well, not anytime soon.
Speaker 3 (26:22):
I don't think TSMC's as you mentioned, of the outset
ninety percent or so market share and high end chips
for many of the key designers of chips, TSMC produces
all of their high end chips. That's true for Apple,
that's true for Nvidia, that's true right now for Qualcomm's
high end chips, for AMD. It's this absolutely central producer
(26:44):
and so deeply internit into the rest of the tech sector,
into how Silicon Valley functions. It's just very hard to
see it being displaced, and rather than being displaced, I
think it's more plausible that we end up in a
situation in ten years where it's really the only player
left stand at the cutting edge because it's two main competitors,
Intel and Samsung are both facing real challenges to their
(27:07):
business of manufacturing high end processor chips.
Speaker 1 (27:11):
How did we get here? Like, how did TSMCA become
this giant?
Speaker 4 (27:15):
Yeah, it's an extraordinary story.
Speaker 3 (27:16):
I mean it really goes back to one individual and
a unique business model. So Morris Chang, the founder of TSMC,
spent most of his early career in the United States,
but was passed over for the CEO job at Texas
Instruments and was approached by the Taiwanese government in the
mid eighties, where he made a number of professional contacts
and they said, would you like to start a business
(27:38):
in Taiwan And he said, yes, but only if I
can try a new business model, which was called the
foundry business model. The intuition was that if you didn't
design any chips, you only manufactured them, you could become
a manufacturing partner for multiple different chip design companies. Therefore
you could build up scale drive down costs but also
improve your manufacturing process because you were gathering data and
(28:00):
know how across a larger number of chips. And that
business model was radical at the time. Most people thought
it would fail. At the time, Morris Chang went to
the leadership at Intel and other titans of the chip
industry for investment. Most of them turned him down, but
it turned out it was the right business model. And
so today TSMC is both the most advanced and the
(28:20):
largest producer of processor chips, and those dynamics are deeply
interlink because it has more data about chip manufacturing, so
it can hone its production processes to a much greater
degree than its competitors can.
Speaker 2 (28:34):
And how important is Morris Chang to TSMC now? The
reason why I'm saying is that he is ninety four
years old. There's going to be a time when TSMC
lives without Morris Chang. Will this interrupt the business at all?
Do you think?
Speaker 1 (28:48):
Well?
Speaker 3 (28:48):
I think TSMC is now and it's a second leader
post Morris Cheng, so it's had multiple leadership transitions. Now,
of course, it's still a company that has in the
past been profoundly influenced by the sets of principles. The
operating model the culture if you will, that he set up.
But I think it's operated independently of his leadership since
(29:11):
he's been retired now for some time. So do I
think it can continue, Yes, but what's continuing as an
organization that he set up with the culture that he
established in a business model that he also deserves credit for,
and for now it looks like both that culture and
that business model are working extraordinarily well.
Speaker 1 (29:30):
We've touched on some regions outside of Taiwan and China
in the US, you know, there's also Japan, South Korea.
Are there ways that countries can kind of take advantage
of the Chip War? Do you see scope for new alliances,
potentially jee political alliances as a result of this.
Speaker 3 (29:50):
I think if you ask most other countries, they will
describe themselves as having been taken advantage of.
Speaker 4 (29:56):
In the context of the Chip.
Speaker 3 (29:58):
War, the metaphor of when elephants fight, the grass gets trampled,
that type of thing is often cited.
Speaker 4 (30:04):
I think there are those aspects of both.
Speaker 3 (30:07):
Certainly, it's the case that for medium sized countries Korea,
Southeast Asian countries, India, they've got to be responsive to
the decisions that both China and the US are making
You can't take either one on head on, either from
a business perspective or from a geopolitical perspective. But I
think also there's a lot of companies that have realized
the industry is going to shift a lot. Already has
(30:28):
shifted a bit, and it's going to shift a lot
more over the coming years, and that will open up opportunities.
I think India is an interesting example. India until recently
played an absolutely marginal role in electronics manufacturing, including in
ship assembly and chip production. But India is looking at
the desire for diversifying supply chains away from China and
(30:48):
has attracted some big US companies as well as some
big Taiwanese companies to set up manufacturing facilities in India
as part of that diversification play. So there's an example
of a country trying to capitalize on the US China rivalry.
But for every example I could give you of a
country capitalizing and give you another example on the opposite
side of a country feeling it was getting squeezed between
(31:10):
two larger powers in a struggle over which it had
no control.
Speaker 1 (31:15):
How are they getting squeezed? I mean, what does that
look like.
Speaker 3 (31:20):
Well, I'll give you an example with Korea. So you know,
Korea is getting squeezed in a couple of different ways.
On the one hand, Korean firms have to follow US
export controls that impact China because every advanced Korean chip
is made partly with tools that are imported from the
United States, and therefore the US claims export control jurisdiction
on basically every ship made in Korea, basically every shipman globally.
(31:43):
And so for both Samsung and Skhiinex, the two leading
Korean shipmakers, they've had real impacts on both the facilities
they currently operate in China, which can't upgrade with the
types of tools they would like to because those tools
are now illegal to ship to China. But also certain
types of memory chips that are used in AI servers
they now can't sell to China, and so that's locked
(32:04):
them out of some potential revenue. It's just like in
Nvidio has been locked out of some potential revenue in China.
So that's how the US has impacted Korean firms. On
the flip side, if you ask yourself, who's the one
of the key potential losers from China self efficiency drive.
Speaker 4 (32:20):
Well, it's Korean firms.
Speaker 3 (32:21):
Korean firms e support large volumes of chips to China,
especially in the memory category, and that's where Chinese firms
like CXMT and YMTC have made really impressive strides at
catching up technologically and displacing Korean firms from at least
the low and mid range of the memory markets. You've
seen already really substantial price impacts in the memory market,
(32:43):
pushing down margins for Samsung and eskhaiinix because of this
increased competition from Chinese firms. And so that's how Chinese
industrial policy is also squeezing companies that previously had access
to the Chinese market in theory still have access, but
in reality are being displaced by Chinese firms.
Speaker 1 (33:01):
How big of a part of the trade talks do
you think this issue is? Because we just had, you know,
the reciprocal tariffs announced on much of the world, we've
also seen more attempts by the Trump administration to kind
of target China's manufacturing kind of throughout the supply chain,
and also this idea of secondary tariffs coming up as
(33:23):
well as well as tariffs on transshipt goods, which seems
to be. They haven't said China in relation to that,
but we can guess it's probably China. So how do
you think that's coming up in trade talks? And do
you think tariffs maybe act counter to like it disincentivizes
countries in a way from targeting this, You.
Speaker 3 (33:43):
Know, I think it's it's really complicated based on which
country one is talking about, and because the tariffs are
still all in flux, it's difficult to draw I think,
confiding conclusions. We still don't know, for example, what the
US is going to do on chip specific tariffs.
Speaker 4 (34:01):
Right now, there are basically no.
Speaker 3 (34:02):
Tariffs on ships, with the exception of some old tariffs
on Chinese ships. But President Trump as repeatedly says there
will be. And so where exactly that lands we don't
yet know. Is it going to be only on China,
It's going to be on the entire world, is going
to be ten percent, fifteen percent, fifty percent, who knows?
And those numbers, those details really really matter. And so
I think we're still as of early August in two
(34:24):
studtel territory just because some of the key tariff decisions
have yet.
Speaker 4 (34:29):
To be made.
Speaker 3 (34:30):
One example, right now, we've got US tariffs on Taiwan
that have been announced. Those could change, but ships aren't
part of that because there's a separate process happening in
the US, And of course, the most important decision that
the US will make visa the trade of Taiwan is
the potential tariff on ships being discussed. So I think
we are in two studitel territory. I think the one
(34:50):
other dam I that's interesting though, I think to me
is that when it comes to the question of Chinese
content in other countries' supply chains, there's an interesting alignment
of interest between what the trumpministration says that it wants
and what businesses in many third countries want, which is
they want more value produced domestically rather than by Chinese companies.
(35:13):
So you go to Vietnam, for example, and none of
the business leaders in Vietnam want to be transhipping goods
from China to the US. They want to be making
the goods and capturing the value themselves.
Speaker 4 (35:25):
Now do they know how to do it? Not always?
Speaker 3 (35:27):
Is it easy to do not necessarily, But there's actually
an interesting alignment between what Trump Minstration vis will say
they want and what the Vietnamese business class would like
to be doing. And so whether or not the Trump
ministration has the diplomatic finesse to actually make that potential
alignment into a reality, I'm not sure, but I think
(35:48):
it is true to note that it's not only the
trumpministration that has been looking nervously at the expansion of
Chinese companies market share across the region and perceiving as
much throughouts Upper Unity.
Speaker 2 (36:01):
Chris, I know you get this question all the time,
and I read the book and I think I know
what the answer is.
Speaker 4 (36:06):
But a lot of people.
Speaker 2 (36:07):
Are probably thinking, you know, if China, for example, does
take over Taiwan and none of the factories of TSMC
are damaged, does that mean that China dominates the semiconductor
manufacturing space.
Speaker 3 (36:23):
I think the first thing to note is that, you know,
the factories themselves are only a small part of what
you need to actually make the advanced ships. You need
the people, critical input, you need materials flowing from Japan.
Speaker 4 (36:35):
You need the.
Speaker 3 (36:36):
Spare parts and software updates flowing from the US and
the Netherlands and elsewhere. So you've got to think about
the entire ecosystem. But in that scenario where you've got
a takeover without any disruption to the chip supply chain.
It would be an extraordinary asset to China. Right now,
the US can tell TSMC don't sell your most advanced
(36:56):
ships for phones or for PCs or for AI to
certain Chinese firms, which means sell them to US firms
or Korean or European firms, and Chinese firms are a disadvantage. Well,
that would flip if China were to take control over Taiwan.
Speaker 4 (37:11):
It could say.
Speaker 3 (37:12):
TSMC provide your most events capabilities to Huawei and none
to Apple, or your most Huawei for GPUs and none
to in Nvidia, And that computing advantage I think would
be fundamentally important, especially as AI relies more and more
on vast volumes of computing power.
Speaker 1 (37:30):
So we've talked about the US export controls. One thing
we haven't talked about is how Chinese companies are avoiding them.
So somehow there are still factories in China and in
Southeast Asia that have chips that are not supposed to
be there from the US. How are they able to
do this?
Speaker 3 (37:50):
Yeah, you know, I think that the key with assessing
chip export controls is that if you want a high
end AI data center, you need tens of thousands of chips,
so we shouldn't ignore when there's one or five or ten,
but they're not strategically relevant. The key question is how
can you get access to tens of thousands of cutting
(38:11):
EDGI chips? And the best evidence we have is that
this is mostly not happening by smuggling in.
Speaker 4 (38:18):
Tens of thousands.
Speaker 3 (38:20):
It's mostly happening by accessing data centers abroad in places
like Malaysia and Thailand and perhaps elsewhere in Southeast Asia
that can legally buy.
Speaker 4 (38:32):
Large volunes of chips.
Speaker 3 (38:34):
So if I were the US government think about this problem,
I would say, of course I want to crack down
on the ten ships here, ten ships there, But you
really want to find are the ten thousand ships and
where they're flowing. And it isn't obvious that we have
chips flowing by the ten thousand increments into Chinese data centers,
but it is obvious that there are chips flowing by
(38:55):
increments that large and larger to countries in Southeast Asia
that don't have domestic demand for that volume of chips.
And so I think when you think of smuggling, you
should think, yes, some people putting an extra chip in
their suitcase and flying it across the border. But that's
probably the less important challenge than the question of how
can you confirm that a data center built in a
(39:18):
third country isn't providing the compute to be in cloud
computing services to a Chinese customer you are hoping to
cut off. And that's part of the debate that is
a bit more technologically complex and therefore not discussed, but
probably more significant in the long run.
Speaker 1 (39:34):
It makes you such interesting stories though. You know, the
men with suitcases full of chips boarding your next axetally happening.
Speaker 2 (39:41):
So the students caught in Dubai with chips.
Speaker 3 (39:44):
And that's why you know the US is you mentioned
the outset Congress is considering legislation to require location verification
for AI data center chips precisely with that in mind.
Speaker 4 (39:57):
You know right now there's a single.
Speaker 3 (40:00):
Sport Control verification official posted by the US government and
all of Southeast Asia, and that single individual is expected
to track the flow of thousands of chips.
Speaker 4 (40:08):
It's just a hopeless task.
Speaker 3 (40:09):
And so you need if you want to actually address
the problem of physical smuggling, you need some sort of
technology to enable it.
Speaker 1 (40:16):
That's fascinating. I mean that person must be traveling quite.
Speaker 4 (40:20):
A bit or busy person. Yes, hard chat.
Speaker 1 (40:23):
So one thing we haven't talked about is rare earths.
They're crucial to semiconductors and AI data centers, and of
course it's an area that China dominates as well. So
I wonder can the US hope to become or can
any country become a leader in chip semiconductors without having
a rare earth's monopoly or at least having a larger
(40:44):
share of the world pie of rare earths.
Speaker 3 (40:47):
Yeah, I think it depends a lot on which specific
mineral you're looking at, So that rare earth magnets are
mostly used in chip making equipment, not chips themselves.
Speaker 4 (40:57):
There's a bunch of other materials.
Speaker 3 (40:59):
Tungsten, gallium, germanium used in the chip making process, which
are also largely processed in China. For a lot of these,
what you find is that the chip industries demand is
a small share of overall demand. So for rare earth
magnets this is the case. There's a lot more demand
that goes to the auto industry than the chip industry.
My sense is that most of the companies in the
(41:20):
chip industry that used rare earth magnets have than a
fair amount of stockpiling in contrast to the auto industry,
and so for the magnet side, I think the industry
is better insulated than the auto industry certainly was in
response to China's shut offs. But for other materials like
gallium germanium, it's less clear. There's not good data what
(41:41):
exactly the disruption might be if China were to impose
comparable restrictions on gallium and germanium, and it hasn't really
yet in terms of the severity. It's threatened restrictions, but
they haven't been biting yet, and so I don't know,
and I don't really think the Chinese government's sure if
it were to try to do the same thing for
these other materials, how impactful or not would it be.
(42:02):
The key question for all these is first how easy
is it or hard is it to do without a
given material? And second how quickly could you bring online
alternative supply and sort of like for rare er it's
not actually rare. They're just polluting and capital intensive to produce,
and that's why they've concentrated in China. The same with
gallium and Germanium. It's not geology that dictates that these
(42:23):
are in China, it's the economics of the mining and
processing industry, which means that these are probably much less
solid monopolies. If you will, then is the Netherlands monopoly
on UV lithography tools.
Speaker 2 (42:36):
So Chris. Earlier this year, China had its Deep Seek moment.
It came out with this LLM that really challenged the
status quo. How impactful was this too, you know, like
the whole AI system, I.
Speaker 4 (42:49):
Think it was very impactful.
Speaker 3 (42:51):
It represented a shift in the axes of progress in AI.
Speaker 4 (42:58):
So for the most period.
Speaker 3 (43:00):
From chatchipt up until the start of this year, advances
in AI were largely driven by what's called pre training,
so training models on larger and larger quantities of data.
And Deep Seek as well as the one reasoning model
from open AI that was released in twenty twenty four,
represented a shift towards reasoning after a question is asked.
(43:22):
So rather than training your model in large quantities of data,
you still do grint and large, but you're you're focusing
more on spending your compute resources on the thinking about
the answer. And this really matters because it has very
different implications for how you use compute and where you
position your compute. So deep Seek they predominantly open source
(43:43):
their model, which means that most queries to deep seak
models take place not in deep seak servers but on
the devices that deep Seak models are run on, so
you can download their model computer use in your phone,
and this has led to its rapid spread both in
China but also globally. But it also poses an interesting
business model question. If everyone's downloading your model, they don't
(44:04):
need you after they downloaded the model, and the model
downloads for free, and so you've now had a number
of Chinese companies Alibaba's quin Models series for example, and
others that have been very permissive in terms of open
sourcing and downloading, and as a result haven't built businesses
despite having really massive distribution of their models. The US firms,
(44:27):
with the exception of Meta, have taken an opposite approach.
They've kept mostly closed source, which means that customers need
to call back to open AI servers every time you
want check sipet to ask you a question, and so
there's very different business models system from this. But there's
also implications for the types of computing controls that we've
been discussing earlier. The initial computing controls were intending or
(44:51):
hoping they could slow the pre training process in China,
but then everyone's pre training process slowed for technological reasons,
and so instead what we've seen is that the controls
have actually been the most impactful on the ability of
Chinese firms to build out their own inference infrastructure. So
you actually saw after deep Seek released its model It's
(45:12):
our one model in January, they had to cut off
access to using their servers to run their model because
they didn't have enough servers to meet the demand of
the entire world. And so they've actually been pushed in
the direction of open sourcing more, I think by the controls,
which had a contradictory and interesting effect of encouraging use
(45:36):
because it is open source, but also undermining the long
term business viability because right now it's unclear what the
funding mechanism will be for either deep Seek or the
other open source models in China in the long run.
Speaker 1 (45:50):
Fascating discussions. Thanks so much for joining us today, Chris.
Speaker 4 (45:53):
Well, thank you for having me.
Speaker 1 (45:55):
You've been listening to Asia Centric from Bloomberg Intelligence and
Kai Di Dimitriva on Hong.
Speaker 2 (45:59):
Kong, also in Hong Kong. You can listen to all
our episodes on Apple Podcasts, Spotify or review listen and
this podcast was also produced and edited by Clara Chen.
Thanks for listening.
Semiconductors are at the heart of America's geopolitical rivalry with China.
They're vital to products we use every day, including smartphones
and computers, and for advanced technologies like AI, electric vehicles
and defense systems. China's made in twenty twenty five policy
a decade ago called for aggressive manufacturing and self sufficiency
(00:23):
in this space.
Speaker 2 (00:25):
That's been worrying Washington ever since, with both the Biden
and Trump administrations adding tougher rules like cutting access to
American technology and certain semiconductors, and it's been met with
limited success. Chinese firms are still finding ways of getting
the hands on US and Taiwanese advanced semiconductors, and the
Chinese government continues to double down on its own industry.
Speaker 1 (00:49):
So where are we in a great chip war? Who's winning?
If anyone? Could we see more thawing in the relationship
as China and the US try to negotiate a trade deal,
and how can regions like Taiwan, South Korea and Japan
benefit from this rivalry. You're listening to Asia Centric from
Bloomberg Intelligence. I'm Katydmitreva in Hong Kong.
Speaker 2 (01:11):
I'm John Lee also in Hong Kong.
Speaker 1 (01:13):
And today we have Chris Miller, who wrote the book
on Justice. He's a professor at Toft University and the
author of Chip War, The Fight for the World's Most
Critical Technology, often seen as required reading for understanding this space.
He joins us in Boston. Thank you for joining us, Chris,
Thank you for having me.
Speaker 2 (01:34):
Chris in your own words, why are semiconductors so important
to the great powers?
Speaker 3 (01:40):
Well, I think there are two reasons why. The first
is that chips are just important for the global economy
writ large. If you look at the world's biggest companies
by market capitalization, they're all either designers of chips, producers
of chips, or companies that really couldn't function without chips.
Speaker 4 (02:00):
Look at global trade.
Speaker 3 (02:01):
The largest single flow in all of global trade is
the flow of chips into China each year, from Taiwan,
from Korea, from Southeast Asia. And so you can't understand
modern economies without chips. That's one reason. The second is
that when you look forward towards the application of artificial
intelligence to every aspect of the economy, what you find
(02:22):
here too, chips are central the advances that we've seen
in AI over the past couple of years have been
driven by better and better semiconductors, which are both used
to train these ultra capable AI systems but also in
deploying them at scale. And so, just like all of
the world's governments, tech CEOs too are betting that if
(02:42):
you want to be able to build and deploy the
best AI systems, you've got to be able to access
to large numbers of high quality AI chips. And so
these two factors, the macroeconomic centrality and the importance to AI,
have put chips not just on the radar screens of
Silicon Valley leaders, but also of the world's top political officials.
Speaker 1 (03:03):
And just sort of a historical point for those who
haven't read the book yet, when did this all begin?
In terms of this importance of chips, Well.
Speaker 3 (03:13):
In some ways it dates back to the origins of
the chip industry. The first ships were invented in the
late nineteen fifties. The aim was to miniatureize computing powers.
You could fit a computer in the nose cone of
a missile and guide it more accurately towards its target,
and this was critical for the Cold War competition between
the United States and the Soviet Union. But it really
(03:35):
kind of kicked up a notch about ten years ago
when the United States was the dominant player in the
chip supply chain, not the dominant manufacturer of chips, but
the dominant producer of many of the key inputs into chips,
the ip the design know how, many of the key tools,
and China realized that it's economy was becoming more and
more dependent on ships that at the time it was
(03:58):
almost completely an importer of. It produce many at all
domestically in the chips that it did produce were all
low end, and so in twenty fourteen, China launched a
major industrial policy effort trying to develop more self sufficiency
in semiconductors. It set up a very large central government
investment fund, and each of the provinces set up their
(04:20):
own investment funds to try to build a self sufficient industry.
And this was seen not just by the US but
also by other key players Taiwan, Korea, Japan is a
economic and potentially down the road, a security threat.
Speaker 2 (04:34):
And where is China now in its aims to become
self sufficient? How would you grade the efforts so far?
Speaker 3 (04:42):
Well, the chip industry is large and complex, and you
get a different answer depending on which corner of the
hip industry.
Speaker 4 (04:48):
You look at. So there's different types of chips.
Speaker 3 (04:51):
There's processor chips, there's memory ships, and there's also different
segments of the supply chain, the design process, the manufacturing itself,
the production of the tool rules that are used to
make chips, and you need to kind of look at
each of these segments individually to really understand how China
stacks up against other competitors. At a high level, I
would say this, On the one hand, China's import of
(05:14):
chips from the rest of the world, which means largely
from Taiwan and Korea and their assembly facilities in Southeast Asia,
generally hits a new high each year it's grown for
most of the last decades. China is importing larger numbers
at least larger dollar values, of chips from abroad. But
China is also producing more at home as well, and
so when it comes to lower range and mid range ships,
(05:36):
China's building a lot of self sufficiency in the types
of ships that go into cars and to refrigerators and
industrial equipment, but the high end ships are getting even
more important. Both the chips that go in phones and PCs,
but also critically the processors and the memory ships that
make AI possible in data centers, and for these cutting
edge ships, China is still mostly importing the chips that
(06:00):
its tech firms need to deploy advanced computing at scale.
Speaker 1 (06:05):
So how far behind is China behind global leaders in
terms of being self sufficient?
Speaker 3 (06:13):
Well, it really depends again on what exactly one wants
to measure. If you ask how far behind is China
when it comes to designing smartphone processors, I would say
it's hardly behind at all. The smartphone processors that Huawei's
high silicon arm designed are by a lot of metrics,
almost as good or just as good as what Apple
(06:33):
or Qualcom would design, And for every different category of
chip you get a different rough metric. But there's a
couple places where China remains meaningfully behind, and it's been
struggling to replicate the capabilities of industry leaders. And I'll
flag two of them which are the most critical. The
first is in the production of the machines that make chips.
(06:56):
If you go inside a chip making facility in China today,
to the complex machines, especially at the high end, are
imported imported from Japan, from the Netherlands and from US firms,
and China has been trying to build out its own
suite of machines that can make chips. But this is
extraordinarily complex equipment, and it's been struggling to make a
(07:17):
lot of progress in replacing foreign chip making tools with
domestic variants. And because of that, China's also struggled to
ramp up cutting edge capacity to manufacture chips itself. And
so you can look, for example, at the leading Chinese firm, SMIC,
which is a chip maker, and compare it to the
leading Taiwanese firm, which is the world's leader, t SMC,
(07:39):
and they have similar manufacturing trajectories. And so you can
look at what SMIC does this year and look at
what TSMC did several years ago and measure the gap
between them as a way of assessing where China stacks up.
And here you find that SMICK right now in Shanghai
has got a seven nanimeter process that it has just
(08:00):
ramped at scale. It's producing ai chips, producing smartphone processors.
But TSMC and Taiwan did that six years ago. So
there's a five or six year gap between what Smith
can produce at scale and what TSMC can produce a scale.
So those are two segments where it's clear there is
still a significant gap between China's capabilities and the leading
Taiwanese US Japanese firms. But those are only two ways
(08:24):
you could measure the gap. And in other segments, especially design,
China's got very impressive capabilities, and I would argue it
has been closing the gap in recent years.
Speaker 1 (08:36):
Maybe a technical point, but you mentioned that China kind
of legs behind in the equipment that makes chips. So
could you describe just a bit why these machines are
so complex and why China is perhaps behind in it.
Because when you think of China as a manufacturer, I
mean the world's factory floor, it's really got the supply
(08:56):
chains down. So what is it about the equipment that
they're not able to do now?
Speaker 4 (09:00):
Yeah, it's a great question.
Speaker 3 (09:02):
When I first myself began studying the industry, it was
kind of a puzzle, why could Chiny to make everything
except for these tools? And the answer, it turns out,
is that these are the most complex and precise tools
humans have literally ever produced. So if you take, for example,
the most complicated and expensive tool called an extreme ultraviolet
photo lithography tool. These are tools produced by just a
(09:24):
single company in the world, ASML of the Netherlands. They
have hundreds of thousands of components, including the flattest mirrors
humans have ever made, the most powerful lasers ever deployed
in the commercial device, the capable of manufacturing at nanometer
scale that's billionths of a meter with close to perfect accuracy.
And that's just one of the tools you need to
(09:44):
make chips. There are many others that are only scarcely
less complex, and so when you start to look at
these tools, you realize that there's nothing that comes close
in terms of complexity and really any other industry. And
so it's not a surprise that China hasn't caught up
because most of these types of tools, there's just one
or two companies in the entire world that can undertake
(10:06):
this activity. And people have been ask why hasn't China
cut up and the production of extreme multi violet photo
lithography tools, Well, no one else has caught up either.
There's just one company that can do it. Others have
tried and failed because the engineering is so complicated and
the business model is so hard to make work, and
so it really is the fact that we're at the
(10:26):
frontiers of engineering the frontiers of physics that explain why
it's not easy for China to replicate this, even though
China's a course, very good at manufacturing many other types
of machinery.
Speaker 2 (10:38):
Asia Centric is produced by Bloomberg Intelligence. We're more than
five hundred experienced analysts and strategists work around the clock
to bring you timely, world class research. Our coverage spans
two hundred market industries, currencies, commodities, and industries, as well
as over two thousand equities and credits. To learn more
about Bloomberg Intelligence, visit b I go on the blue
(11:00):
the terminal. If you like what you hear, don't forget
to subscribe and chair Chris, Look, I read your book,
But for the uninitiated who didn't read your book, can
you explain why like this industry lends itself to monopolies
like you mentioned that ASML pretty much like no one
else makes these you know, lithic of methography on machines,
(11:24):
and I think you said no one else makes these
TSMC also manufactures over ninety percent of the advanced AI chips.
Why is this the case? That there are monopolies formed
in the supply chain.
Speaker 1 (11:35):
Yeah.
Speaker 3 (11:35):
I think there are two factors that lend the chip
industry to these sort of oligopolistic rin monopolistic markets. The
first is the capital expenditures are often tremendous. So for TSMC,
every new factory they build is the world's most expensive
factory because each factory is more expensive and breaks the
(11:56):
prior record. We're talking twenty or twenty five billion dollars
per FACI, and so this scope for a new startup
firm to compete is approximately zero. That has really kept
competition down because the number of players that could even
conceive of taking on TSMC is very small, and the
(12:16):
same dynamic is true in other segments of the supply chain.
Speaker 4 (12:19):
The dollar values are lower, but not much lower.
Speaker 3 (12:22):
Even if it's only three billion dollars for a new
factory that makes the ultra purified a silicon wafers, for example,
that's still a monumental capital expense for most companies, and
it deters new entrants from even trying. The second factor
is that the R and D is really extraordinarily highly specialized,
and for a lot of the process steps and making chips,
(12:44):
the only way you learn about how to undertake that
specific step is by having already been in the industry
and learning alongside your suppliers and your customers.
Speaker 4 (12:55):
And so, if you think of.
Speaker 3 (12:56):
The way the chip industry's R and D works, every year,
the entire industry works together to bring a new process
that's more capable than the prior years, often dramatically more
capable in terms of the computing power it produces. And
so you've got really tight alignment between the existing chemicals producers,
existing machinery producers, existing chip manufacturers themselves to make sure
(13:17):
that these processes arrive on time and that all the
necessary components are together. There's a lot of shared learning
inside of that small circle. But if you're an outsider
who's never been part of that learning process, there's no
way to do it on your own. You can't just
test on your own very easily whether your new chemical
will work with a existing extreme ultraviolet photogthography machine. If
(13:40):
you want to buy one of those machines yourself, it
costs you one hundred and fifty million dollars. And moreover,
companies aren't going to let you run a lot of
tests in their fab because they just spend twenty billion
dollars building a fab they.
Speaker 4 (13:50):
Need to run that at full capacity.
Speaker 3 (13:52):
So it's really hard to break in from an R
and D perspective, And when you combine that with the
capital intensivity, it means that many of the firms that
have key roles in the supply chain have had them
for the last decade or in some cases the last
three or four decades, and that has kept competition down
in the hip industry.
Speaker 1 (14:10):
So you wrote your book in twenty twenty two, that
was under the Biden administration. We're now under the Trump
administration two point zero. What are the big differences you've
seen in how both administrations have handled this issue? Are
there some similarities that come up, but also some of
the big big differences, because of course under Biden we
had the Chips IRA Act, But under both administrations it
(14:35):
seems like a bipart is an issue.
Speaker 3 (14:38):
Yeah, I think it is, and I think that the
similarities are to me more substantial than the differences. So first,
there's a focus on domestic manufacturing in the United States,
an effort to reduce reliance both on the lower end
ship produced in China, but also dependence on high end
chips produced in Taiwan the Chips Act. Although Biden claimed
credit for it, it actually emerged under the trumpetministration. That was
(15:00):
when the Congress first started discussing the bill, and people
like Secretary of State Mike Pompeo played a major role
in some of the early investments that the Taiwanese from
TSMC announced the United States. So there's been a through
line of support for manufacturing chips in the United States
that has been bipartisan.
Speaker 4 (15:17):
I think that's a point of continuity.
Speaker 3 (15:20):
I think the second point in continuity has been a desire
to stop China from accessing the most advanced ship making technologies.
The export controls on ASML tools, for example, started under
the first administration, were expanded under Biden, and we'll see
they could well be expanded further at some point under
the current TRUP administration. So that's also a through line.
(15:40):
And I think the third is concern about Chinese industrial
subsidies in the chip industry and implications for trade policy.
The Trumpdministration in round one imposed higher tariffs and Chinese
ships imported the United States. Biden expanded those tariffs, and
of course President Trump right now is considering what he's
going to do.
Speaker 4 (16:01):
On chip tariffs.
Speaker 3 (16:01):
But there's widespread expectation that there will be another round
of chip tariffs coming at some point later this year,
and so that's also a through line of concern about
Chinese subsidies. I think there are differences. Obviously, you know,
we're going to see where exactly Trump's chip tariffs land,
and they might end up being much broader on a
whole range of different countries, not just China focused.
Speaker 4 (16:21):
If so, that would be a difference.
Speaker 3 (16:22):
But I think when you zoom out, you actually end
up seeing a pheromounta continuity, despite some nuances that are different.
Speaker 1 (16:31):
Chris.
Speaker 2 (16:31):
President Trump seems to be dialing back on some of
the policies now. Under the previous administration, they banned the
export of high end in Nvidia chips into China, but
Donald Trump recently suggested that they could allow the export
of certain Age twenty chips back into the country. What
does this mean, Well, I.
Speaker 3 (16:54):
Think it's too soon to have a confident judgment. I
would say when the Biden administration left office, in Videau's
H twenty were still legal to sell China. Then in
month four of the trumpdministration, the President banned the export
of Age twenty chips, and now there's suggestions they might
be allowed again. As of today, we're still at a
tighter restrictive policy than when the Biden administration ended.
Speaker 4 (17:17):
When the Trump administration started.
Speaker 3 (17:18):
We'll see exactly how this plays out, but I think
it speaks again to a continuity, which is that both
Biden and Trump feel like they need to balance action
on chip issues with the broader.
Speaker 4 (17:32):
US China relationship.
Speaker 3 (17:33):
So into the Biden administration, it was regularly the case
that the Commerce Department would have a new regulation ready
that would limit this or that aspect of the chip trade,
and then it would wait for three months or six
months or nine months while the State Department found what
they thought was a quote unquote good window in which
to roll out a new restriction that the Chinese wouldn't like.
And cour we're seeing the trumpministration as something similar.
Speaker 4 (17:55):
Right now.
Speaker 3 (17:56):
The President wants to negotiate a trade deal with China.
Their White House is trying to get rare earth magnets
flowing again, and so there's concern in the White House
that if they were to take punitive steps on chips
or on chip making tools, that this would disrupt the
US China relationship. One can debate whether that's the right approach,
but I think it is a striking similarity to a
(18:18):
lot of the delays and the kind of half hearted
decision making that you saw in the Biden administration when
it came to whether or not to restrict certain types
of tools or chips.
Speaker 1 (18:30):
So, after the Trump administration recently allowed the export of
H twenty chips, very closely afterwards, China called in Nvidia
staff to have a discussion with them behind that is
at least partly the Trump administration suggesting to track these chips.
And this is all happening within the context of a
(18:52):
trade war and trade negotiations. So one of the questions
we've been thinking about is, you know, why are the
US and China right now fighting over what are essentially
kind of third rate chips. I mean, these aren't like
the most advanced chips out there. As John ass like,
how should we read this kind of back and forth.
Speaker 3 (19:13):
Well, I wouldn't say it's right to consider the H
twenties third rate chips. I think if you're a Chinese
tech company, these are first rate. These are the US
you can get at scale can you smuggle in small
volumes of the higher quality H one hundred yes. Can
you access small volumes of Huawei ASSN chips Yes? But
can you access either by the tens or hundreds of thousands, No,
(19:34):
both because Huawei can't produce that efficient scale and because
smuggling you can do a medium scale, but not at
the scale you need. So you know, judging by the
purchasing patterns of byte Edance and Ali Baba Intense, these
are great chips, which is why they really want to
spend hundreds of millions or billions of.
Speaker 4 (19:51):
Dollars on them.
Speaker 3 (19:52):
And I think that that is why this is such
a critical issue, both for the US and for Chinese
tech firms.
Speaker 4 (19:58):
If they really were third rate, we wouldn't be talking
about them.
Speaker 2 (20:00):
Chris like and Vidia CEO Jensen Huang was very critical
about the US ban on chips to China. He said
it hurts US businesses, and he also insinuated that it's
going to accelerate China's domestic semiconductive development. Do you agree
with him? Do you think these mergers are in some
way self defeating?
Speaker 3 (20:20):
Well, I think there's no doubt that in the short
run it reduces in video sales. That's certainly clear. So
does it hurt that business? I think the answer is yes.
One of the dilemmas that any sort of export control
regime faces is that if you restrict the sale of
a certain tool, you hurt the tool makers, but you
benefit the domestic companies that use that tool, who have
(20:42):
less foreign competition. So, for example, if you restrict the
sale of chip making tools, it's bad for the companies
that make chip making tools, but it's actually better for
chip makers in the West, including in Taiwan and South Korea,
because they don't have strong competition in China at the
cutting edge. And if you thought about the counterfactuals pose
the Trudministration had not leaned on the Netherlands in twenty
(21:03):
nineteen and said the Netherlands don't sell UV tools to China,
China would be operating today a fleet of many UV tools,
possibly the second largest behind Taiwan in the world, and
that'd be a totally different landscape for the world's chip makers.
It'd be better for sellers of chip maker tools, but
worse for chip makers. And we're in a similar dynamic
right now with chips. If you sell more chips it's
(21:24):
good for chip makers, but it's probably worse for the
companies that use chips, which are the big tech companies,
cloud computing companies.
Speaker 4 (21:31):
And AI labs. Why is it that.
Speaker 3 (21:35):
We haven't seen large deployments of AI chips buy Chinese
firms outside of China's because they can't produce enough, because
they haven't been able to access large volumes of the
high end tools that foreign firms have. And so I
think that the trade off that the US has to
navigate is, Yes, it's true that it's worse for your
(21:56):
chip makers if you don't allow sales to China, but
it's also probably better for your tech firms that use
those ships because they won't have strong Chinese competitors or
as strong Chinese competitors, and that's a trade off. I
think there's no doubt about that. I think the other
question is is US policy accelerating China's technological development?
Speaker 4 (22:15):
And here I look for empirical data points.
Speaker 3 (22:19):
We know that in twenty fourteen, China launched its first
national level Integrated Circuit Investment Fund.
Speaker 4 (22:26):
It's been re upped now twice.
Speaker 3 (22:28):
Since then at the national level, and each province has
If you ask yourself is it clear from the data,
and the data's kind of messy, but is it clear
that there's been a major increase in Chinese government subsidies
towards the ship industry after export controls? It's not really
clear because it started in twenty fourteen and it's been,
as far as we can tell, roughly steady since then.
(22:50):
If you ask, has there been a noticeable increase in
the Chinese government pushing companies to buy domestic chips after
export controls?
Speaker 4 (22:59):
I don't think it's clear.
Speaker 3 (23:00):
I think you've seen from twenty fourteen to present a
steady push to buy a local And so if you ask,
you know, what's the evidence that there's been a major
uptick in self sufficiency, either support or pressure from the
Chinese government.
Speaker 4 (23:13):
Around export controls? It's hard to find.
Speaker 3 (23:16):
Now, could you argue that domestic Chinese tech firms before
export controls would have said, of course, we're going to
test local products. They would have bought local products and
put them in the closet and instead relied on imported products.
And now if they can't import products, they've got no
choice but to take the domestic products. Seriously, I think
(23:37):
you could argue that and perhaps that's happening to some degree,
but I would say, again, the evidence for that is
not super robust. And so if you're just trying to
be empirical and say, what's been the evident changed in
Chinese policy or the approach of Chinese tech firms or
export controls, I would say it looks pretty continuous over time.
In other words, export controls didn't obviously have a major impact.
Speaker 1 (23:59):
Are the expert restrictions enough though? Are there other things
that you see the administration moving towards well?
Speaker 3 (24:06):
I think a lot depends on the broader US China relationship.
There are certainly voices in the administration that want to
do a lot more on the restrictive side. There are
also voices administration that want to do a lot more
on the promoting US technological advances side. But of course
the administration has many priorities, and it's hard to see
which is going to win out. I think in terms
of the restrictions, it's very clear that the White House
(24:29):
is going to try to balance the Hawk's desire for
restrictions with the broader US channel relationship. And so as
we approach what many people think is going to be
a summit, this fall. I think history would suggest you
shouldn't expect freemen restrictions and then run up to a
summit on the promoting technological development side. I think there
certainly are a lot of voices administration that want to
(24:49):
see more money flowing to new firms with disruptive technology.
But there's also a desire for pretty dramatic budget cuts
to science and technology in the USA, and that cuts
in the opposite direction. So how exactly these political priorities
balance out is hard to predict with any certainty.
Speaker 1 (25:07):
Just a quick follow up on the point you made
about this balance or dichotomy between US producers and exporters
of chips and tech firms, how do you see the
Trump administration to serve the next few years navigating that?
Does it mean that they need to find other ways
of supporting chip makers, for example, if they're going to
be closing the route to China.
Speaker 3 (25:30):
Well, I don't think any administration has done a great
answer to that question. I mean, China's obviously the second
large economy in the world, and you know, twenty percent
of GDP is a really large market to not be
able to serve. I think the US has tried to
argue to tech firms. Look, you can't serve this twenty percent.
You have the rest of the eighty percent you can serve.
And I think tech companies are inclined to serve that
(25:53):
other eighty percent anyway to a large degree.
Speaker 4 (25:56):
But is it right that companies.
Speaker 3 (25:58):
Would be more profitable have high revenue if they could
sell to the entire one hundred percent of GDP?
Speaker 4 (26:02):
For sure, no doubt about that, Chris.
Speaker 2 (26:05):
I'd love to talk about TSMC now. The company became
the first Asian stock to reach a market capitalization of
one trillion dollars. That was just you know recently. Can
anyone challenge TSMC's dominance in this space.
Speaker 4 (26:21):
Well, not anytime soon.
Speaker 3 (26:22):
I don't think TSMC's as you mentioned, of the outset
ninety percent or so market share and high end chips
for many of the key designers of chips, TSMC produces
all of their high end chips. That's true for Apple,
that's true for Nvidia, that's true right now for Qualcomm's
high end chips, for AMD. It's this absolutely central producer
(26:44):
and so deeply internit into the rest of the tech sector,
into how Silicon Valley functions. It's just very hard to
see it being displaced, and rather than being displaced, I
think it's more plausible that we end up in a
situation in ten years where it's really the only player
left stand at the cutting edge because it's two main competitors,
Intel and Samsung are both facing real challenges to their
(27:07):
business of manufacturing high end processor chips.
Speaker 1 (27:11):
How did we get here? Like, how did TSMCA become
this giant?
Speaker 4 (27:15):
Yeah, it's an extraordinary story.
Speaker 3 (27:16):
I mean it really goes back to one individual and
a unique business model. So Morris Chang, the founder of TSMC,
spent most of his early career in the United States,
but was passed over for the CEO job at Texas
Instruments and was approached by the Taiwanese government in the
mid eighties, where he made a number of professional contacts
and they said, would you like to start a business
(27:38):
in Taiwan And he said, yes, but only if I
can try a new business model, which was called the
foundry business model. The intuition was that if you didn't
design any chips, you only manufactured them, you could become
a manufacturing partner for multiple different chip design companies. Therefore
you could build up scale drive down costs but also
improve your manufacturing process because you were gathering data and
(28:00):
know how across a larger number of chips. And that
business model was radical at the time. Most people thought
it would fail. At the time, Morris Chang went to
the leadership at Intel and other titans of the chip
industry for investment. Most of them turned him down, but
it turned out it was the right business model. And
so today TSMC is both the most advanced and the
(28:20):
largest producer of processor chips, and those dynamics are deeply
interlink because it has more data about chip manufacturing, so
it can hone its production processes to a much greater
degree than its competitors can.
Speaker 2 (28:34):
And how important is Morris Chang to TSMC now? The
reason why I'm saying is that he is ninety four
years old. There's going to be a time when TSMC
lives without Morris Chang. Will this interrupt the business at all?
Do you think?
Speaker 1 (28:48):
Well?
Speaker 3 (28:48):
I think TSMC is now and it's a second leader
post Morris Cheng, so it's had multiple leadership transitions. Now,
of course, it's still a company that has in the
past been profoundly influenced by the sets of principles. The
operating model the culture if you will, that he set up.
But I think it's operated independently of his leadership since
(29:11):
he's been retired now for some time. So do I
think it can continue, Yes, but what's continuing as an
organization that he set up with the culture that he
established in a business model that he also deserves credit for,
and for now it looks like both that culture and
that business model are working extraordinarily well.
Speaker 1 (29:30):
We've touched on some regions outside of Taiwan and China
in the US, you know, there's also Japan, South Korea.
Are there ways that countries can kind of take advantage
of the Chip War? Do you see scope for new alliances,
potentially jee political alliances as a result of this.
Speaker 3 (29:50):
I think if you ask most other countries, they will
describe themselves as having been taken advantage of.
Speaker 4 (29:56):
In the context of the Chip.
Speaker 3 (29:58):
War, the metaphor of when elephants fight, the grass gets trampled,
that type of thing is often cited.
Speaker 4 (30:04):
I think there are those aspects of both.
Speaker 3 (30:07):
Certainly, it's the case that for medium sized countries Korea,
Southeast Asian countries, India, they've got to be responsive to
the decisions that both China and the US are making
You can't take either one on head on, either from
a business perspective or from a geopolitical perspective. But I
think also there's a lot of companies that have realized
the industry is going to shift a lot. Already has
(30:28):
shifted a bit, and it's going to shift a lot
more over the coming years, and that will open up opportunities.
I think India is an interesting example. India until recently
played an absolutely marginal role in electronics manufacturing, including in
ship assembly and chip production. But India is looking at
the desire for diversifying supply chains away from China and
(30:48):
has attracted some big US companies as well as some
big Taiwanese companies to set up manufacturing facilities in India
as part of that diversification play. So there's an example
of a country trying to capitalize on the US China rivalry.
But for every example I could give you of a
country capitalizing and give you another example on the opposite
side of a country feeling it was getting squeezed between
(31:10):
two larger powers in a struggle over which it had
no control.
Speaker 1 (31:15):
How are they getting squeezed? I mean, what does that
look like.
Speaker 3 (31:20):
Well, I'll give you an example with Korea. So you know,
Korea is getting squeezed in a couple of different ways.
On the one hand, Korean firms have to follow US
export controls that impact China because every advanced Korean chip
is made partly with tools that are imported from the
United States, and therefore the US claims export control jurisdiction
on basically every ship made in Korea, basically every shipman globally.
(31:43):
And so for both Samsung and Skhiinex, the two leading
Korean shipmakers, they've had real impacts on both the facilities
they currently operate in China, which can't upgrade with the
types of tools they would like to because those tools
are now illegal to ship to China. But also certain
types of memory chips that are used in AI servers
they now can't sell to China, and so that's locked
(32:04):
them out of some potential revenue. It's just like in
Nvidio has been locked out of some potential revenue in China.
So that's how the US has impacted Korean firms. On
the flip side, if you ask yourself, who's the one
of the key potential losers from China self efficiency drive.
Speaker 4 (32:20):
Well, it's Korean firms.
Speaker 3 (32:21):
Korean firms e support large volumes of chips to China,
especially in the memory category, and that's where Chinese firms
like CXMT and YMTC have made really impressive strides at
catching up technologically and displacing Korean firms from at least
the low and mid range of the memory markets. You've
seen already really substantial price impacts in the memory market,
(32:43):
pushing down margins for Samsung and eskhaiinix because of this
increased competition from Chinese firms. And so that's how Chinese
industrial policy is also squeezing companies that previously had access
to the Chinese market in theory still have access, but
in reality are being displaced by Chinese firms.
Speaker 1 (33:01):
How big of a part of the trade talks do
you think this issue is? Because we just had, you know,
the reciprocal tariffs announced on much of the world, we've
also seen more attempts by the Trump administration to kind
of target China's manufacturing kind of throughout the supply chain,
and also this idea of secondary tariffs coming up as
(33:23):
well as well as tariffs on transshipt goods, which seems
to be. They haven't said China in relation to that,
but we can guess it's probably China. So how do
you think that's coming up in trade talks? And do
you think tariffs maybe act counter to like it disincentivizes
countries in a way from targeting this, You.
Speaker 3 (33:43):
Know, I think it's it's really complicated based on which
country one is talking about, and because the tariffs are
still all in flux, it's difficult to draw I think,
confiding conclusions. We still don't know, for example, what the
US is going to do on chip specific tariffs.
Speaker 4 (34:01):
Right now, there are basically no.
Speaker 3 (34:02):
Tariffs on ships, with the exception of some old tariffs
on Chinese ships. But President Trump as repeatedly says there
will be. And so where exactly that lands we don't
yet know. Is it going to be only on China,
It's going to be on the entire world, is going
to be ten percent, fifteen percent, fifty percent, who knows?
And those numbers, those details really really matter. And so
I think we're still as of early August in two
(34:24):
studtel territory just because some of the key tariff decisions
have yet.
Speaker 4 (34:29):
To be made.
Speaker 3 (34:30):
One example, right now, we've got US tariffs on Taiwan
that have been announced. Those could change, but ships aren't
part of that because there's a separate process happening in
the US, And of course, the most important decision that
the US will make visa the trade of Taiwan is
the potential tariff on ships being discussed. So I think
we are in two studitel territory. I think the one
(34:50):
other dam I that's interesting though, I think to me
is that when it comes to the question of Chinese
content in other countries' supply chains, there's an interesting alignment
of interest between what the trumpministration says that it wants
and what businesses in many third countries want, which is
they want more value produced domestically rather than by Chinese companies.
(35:13):
So you go to Vietnam, for example, and none of
the business leaders in Vietnam want to be transhipping goods
from China to the US. They want to be making
the goods and capturing the value themselves.
Speaker 4 (35:25):
Now do they know how to do it? Not always?
Speaker 3 (35:27):
Is it easy to do not necessarily, But there's actually
an interesting alignment between what Trump Minstration vis will say
they want and what the Vietnamese business class would like
to be doing. And so whether or not the Trump
ministration has the diplomatic finesse to actually make that potential
alignment into a reality, I'm not sure, but I think
(35:48):
it is true to note that it's not only the
trumpministration that has been looking nervously at the expansion of
Chinese companies market share across the region and perceiving as
much throughouts Upper Unity.
Speaker 2 (36:01):
Chris, I know you get this question all the time,
and I read the book and I think I know
what the answer is.
Speaker 4 (36:06):
But a lot of people.
Speaker 2 (36:07):
Are probably thinking, you know, if China, for example, does
take over Taiwan and none of the factories of TSMC
are damaged, does that mean that China dominates the semiconductor
manufacturing space.
Speaker 3 (36:23):
I think the first thing to note is that, you know,
the factories themselves are only a small part of what
you need to actually make the advanced ships. You need
the people, critical input, you need materials flowing from Japan.
Speaker 4 (36:35):
You need the.
Speaker 3 (36:36):
Spare parts and software updates flowing from the US and
the Netherlands and elsewhere. So you've got to think about
the entire ecosystem. But in that scenario where you've got
a takeover without any disruption to the chip supply chain.
It would be an extraordinary asset to China. Right now,
the US can tell TSMC don't sell your most advanced
(36:56):
ships for phones or for PCs or for AI to
certain Chinese firms, which means sell them to US firms
or Korean or European firms, and Chinese firms are a disadvantage. Well,
that would flip if China were to take control over Taiwan.
Speaker 4 (37:11):
It could say.
Speaker 3 (37:12):
TSMC provide your most events capabilities to Huawei and none
to Apple, or your most Huawei for GPUs and none
to in Nvidia, And that computing advantage I think would
be fundamentally important, especially as AI relies more and more
on vast volumes of computing power.
Speaker 1 (37:30):
So we've talked about the US export controls. One thing
we haven't talked about is how Chinese companies are avoiding them.
So somehow there are still factories in China and in
Southeast Asia that have chips that are not supposed to
be there from the US. How are they able to
do this?
Speaker 3 (37:50):
Yeah, you know, I think that the key with assessing
chip export controls is that if you want a high
end AI data center, you need tens of thousands of chips,
so we shouldn't ignore when there's one or five or ten,
but they're not strategically relevant. The key question is how
can you get access to tens of thousands of cutting
(38:11):
EDGI chips? And the best evidence we have is that
this is mostly not happening by smuggling in.
Speaker 4 (38:18):
Tens of thousands.
Speaker 3 (38:20):
It's mostly happening by accessing data centers abroad in places
like Malaysia and Thailand and perhaps elsewhere in Southeast Asia
that can legally buy.
Speaker 4 (38:32):
Large volunes of chips.
Speaker 3 (38:34):
So if I were the US government think about this problem,
I would say, of course I want to crack down
on the ten ships here, ten ships there, But you
really want to find are the ten thousand ships and
where they're flowing. And it isn't obvious that we have
chips flowing by the ten thousand increments into Chinese data centers,
but it is obvious that there are chips flowing by
(38:55):
increments that large and larger to countries in Southeast Asia
that don't have domestic demand for that volume of chips.
And so I think when you think of smuggling, you
should think, yes, some people putting an extra chip in
their suitcase and flying it across the border. But that's
probably the less important challenge than the question of how
can you confirm that a data center built in a
(39:18):
third country isn't providing the compute to be in cloud
computing services to a Chinese customer you are hoping to
cut off. And that's part of the debate that is
a bit more technologically complex and therefore not discussed, but
probably more significant in the long run.
Speaker 1 (39:34):
It makes you such interesting stories though. You know, the
men with suitcases full of chips boarding your next axetally happening.
Speaker 2 (39:41):
So the students caught in Dubai with chips.
Speaker 3 (39:44):
And that's why you know the US is you mentioned
the outset Congress is considering legislation to require location verification
for AI data center chips precisely with that in mind.
Speaker 4 (39:57):
You know right now there's a single.
Speaker 3 (40:00):
Sport Control verification official posted by the US government and
all of Southeast Asia, and that single individual is expected
to track the flow of thousands of chips.
Speaker 4 (40:08):
It's just a hopeless task.
Speaker 3 (40:09):
And so you need if you want to actually address
the problem of physical smuggling, you need some sort of
technology to enable it.
Speaker 1 (40:16):
That's fascinating. I mean that person must be traveling quite.
Speaker 4 (40:20):
A bit or busy person. Yes, hard chat.
Speaker 1 (40:23):
So one thing we haven't talked about is rare earths.
They're crucial to semiconductors and AI data centers, and of
course it's an area that China dominates as well. So
I wonder can the US hope to become or can
any country become a leader in chip semiconductors without having
a rare earth's monopoly or at least having a larger
(40:44):
share of the world pie of rare earths.
Speaker 3 (40:47):
Yeah, I think it depends a lot on which specific
mineral you're looking at, So that rare earth magnets are
mostly used in chip making equipment, not chips themselves.
Speaker 4 (40:57):
There's a bunch of other materials.
Speaker 3 (40:59):
Tungsten, gallium, germanium used in the chip making process, which
are also largely processed in China. For a lot of these,
what you find is that the chip industries demand is
a small share of overall demand. So for rare earth
magnets this is the case. There's a lot more demand
that goes to the auto industry than the chip industry.
My sense is that most of the companies in the
(41:20):
chip industry that used rare earth magnets have than a
fair amount of stockpiling in contrast to the auto industry,
and so for the magnet side, I think the industry
is better insulated than the auto industry certainly was in
response to China's shut offs. But for other materials like
gallium germanium, it's less clear. There's not good data what
(41:41):
exactly the disruption might be if China were to impose
comparable restrictions on gallium and germanium, and it hasn't really
yet in terms of the severity. It's threatened restrictions, but
they haven't been biting yet, and so I don't know,
and I don't really think the Chinese government's sure if
it were to try to do the same thing for
these other materials, how impactful or not would it be.
(42:02):
The key question for all these is first how easy
is it or hard is it to do without a
given material? And second how quickly could you bring online
alternative supply and sort of like for rare er it's
not actually rare. They're just polluting and capital intensive to produce,
and that's why they've concentrated in China. The same with
gallium and Germanium. It's not geology that dictates that these
(42:23):
are in China, it's the economics of the mining and
processing industry, which means that these are probably much less
solid monopolies. If you will, then is the Netherlands monopoly
on UV lithography tools.
Speaker 2 (42:36):
So Chris. Earlier this year, China had its Deep Seek moment.
It came out with this LLM that really challenged the
status quo. How impactful was this too, you know, like
the whole AI system, I.
Speaker 4 (42:49):
Think it was very impactful.
Speaker 3 (42:51):
It represented a shift in the axes of progress in AI.
Speaker 4 (42:58):
So for the most period.
Speaker 3 (43:00):
From chatchipt up until the start of this year, advances
in AI were largely driven by what's called pre training,
so training models on larger and larger quantities of data.
And Deep Seek as well as the one reasoning model
from open AI that was released in twenty twenty four,
represented a shift towards reasoning after a question is asked.
(43:22):
So rather than training your model in large quantities of data,
you still do grint and large, but you're you're focusing
more on spending your compute resources on the thinking about
the answer. And this really matters because it has very
different implications for how you use compute and where you
position your compute. So deep Seek they predominantly open source
(43:43):
their model, which means that most queries to deep seak
models take place not in deep seak servers but on
the devices that deep Seak models are run on, so
you can download their model computer use in your phone,
and this has led to its rapid spread both in
China but also globally. But it also poses an interesting
business model question. If everyone's downloading your model, they don't
(44:04):
need you after they downloaded the model, and the model
downloads for free, and so you've now had a number
of Chinese companies Alibaba's quin Models series for example, and
others that have been very permissive in terms of open
sourcing and downloading, and as a result haven't built businesses
despite having really massive distribution of their models. The US firms,
(44:27):
with the exception of Meta, have taken an opposite approach.
They've kept mostly closed source, which means that customers need
to call back to open AI servers every time you
want check sipet to ask you a question, and so
there's very different business models system from this. But there's
also implications for the types of computing controls that we've
been discussing earlier. The initial computing controls were intending or
(44:51):
hoping they could slow the pre training process in China,
but then everyone's pre training process slowed for technological reasons,
and so instead what we've seen is that the controls
have actually been the most impactful on the ability of
Chinese firms to build out their own inference infrastructure. So
you actually saw after deep Seek released its model It's
(45:12):
our one model in January, they had to cut off
access to using their servers to run their model because
they didn't have enough servers to meet the demand of
the entire world. And so they've actually been pushed in
the direction of open sourcing more, I think by the controls,
which had a contradictory and interesting effect of encouraging use
(45:36):
because it is open source, but also undermining the long
term business viability because right now it's unclear what the
funding mechanism will be for either deep Seek or the
other open source models in China in the long run.
Speaker 1 (45:50):
Fascating discussions. Thanks so much for joining us today, Chris.
Speaker 4 (45:53):
Well, thank you for having me.
Speaker 1 (45:55):
You've been listening to Asia Centric from Bloomberg Intelligence and
Kai Di Dimitriva on Hong.
Speaker 2 (45:59):
Kong, also in Hong Kong. You can listen to all
our episodes on Apple Podcasts, Spotify or review listen and
this podcast was also produced and edited by Clara Chen.
Thanks for listening.
Hosts And Creators
John Lee
Tom Corbett
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!
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.