Despite some reports out there, this isn't domestication of chip making in China at all--they are still using old imported lithography machines, just in a novel way to juice out chips a few nanometers below what they are intended to make. The machines themselves are still from the Netherlands with German mirrors and full of American IP with complexity beyond the limit of human comprehension. What China is finding out now is the limits to their old equipment. It doesn't reflect a failure in sanctions regime at all, and I'd guess that their capacity to keep up with the West in chips will diverge over time. Taiwan was making 7nm chips five years ago.
Sep 14, 2023·edited Sep 14, 2023Liked by Noah Smith
Nice post. People seem to wishful-think sanctions into a magic bullet they've never been, then complain the imaginary bullet isn't lethal.
Sanctions are just a string of roadblocks. They're lousy at stopping things altogether, but pretty good at making things expensive, inconvenient, and a whole lot slower. If "expensive, inconvenient, and a whole lot slower" gives you negotiating leverage, or time to counter an opponent, that's a good case for sanctions.
If you don't really have a proposal worth negotiating, or a plan to put time on your side, sanctions aren't going to fix that for you. Sanctions exist to complement and strengthen the rest of your negotiating strategy or play-for-time strategy. They're not a strategy by themselves.
Personally I think our China strategy makes sense. Sanctions buy us more time for all our other tools to work. It's up to us to use that time well.
Sometimes, manufacturing and commitment outweigh a technology edge. Russia makes more artillery shells than the USA and Europe combined; it's even sending more drones to Ukraine than the west, despite drones being high-tech compared to explosive-filled 155mm cones of steel.
So even Russia, never mind China, is managing their quality gap with a quantity advantage.
Sanctions will buy the USA some time. But our only lasting deterrent will be if America relearns how to be an "arsenal of democracy." Billion-dollar systems built one at a time are a sign of vulnerability, not strength; the USA that'll deter China is a USA that can build military equipment not just with quality, but in quantity.
The world is truly mad. What does a Chinese company producing a consumer phone have anything to do with America? So killing this innovative business seems make you all happier?
To me, this is a reassuring development. Not understanding the technology, I had thought that the export controls was more akin to a commodity cutoff than an important strategic move to slow down China's progress. If China believes it can overcome the export controls in time, then they are less likely to take any drastic actions. Thanks for clarifying!
It's funny that in China, ignorant people like Noah would be seen as products of the bureaucratic system, while in the United States, such people are seen as experts and think tanks.
This post makes some of the same mistakes that many other commentators have on this topic (mistakes that are common on both sides of the issue, btw). Two things can be true: 1/ The 9000s isn't proof that US export controls have no effect; 2/ Huawei/SMIC have made a genuine capability breakthrough on a far faster timeline than was predicted just 2 or so years ago.
When Huawei was first added to the Entity List in 2019 and then restrictions were tightened further in 2020 to preclude its use of TSMC to fab its in-house HiSilicon chips, the state of the art was 7nm (with 5nm in sight). Lots of observation at the time pegged China as being "roughly a decade" away from being able to reach best in class. And yet, 3-4 years later, they've now reached what was (at that time) the best in class.
This has taken them less than half the time many predicted and demonstrates not so much the power of Beijing's industrial policy, but the power of what can be achieved when the incentives of China's leading private sector players (e.g., Huawei) are forcibly aligned with the Chinese state's due to US export controls. Huawei was happy to continue procuring from US suppliers (it was buying ~$20B a year!). Yes, it was developing in-house alternatives for many components as part of its longstanding strategy of vertical integration, but it saw its central role in global tech supply chains as something that was desirable, not regrettable. Until the US government intervened.
The US has now given Huawei no choice but to throw its entire self into developing credible alternatives to US and western suppliers and it's bringing its army of 100,000+ engineers to bear on the problem. This was something Beijing's industrial policy never could have accomplished on its own.
Although it's true that the cat is, to some degree out of the bag here, it would be a mistake to claim that US has no policy choice at this point and must continue on the export control course. One claim in the piece is particularly outlandish:
"It’s hilariously unrealistic to think that if the Biden administration dropped all export controls tomorrow, Chinese companies would just go right back to buying their chips from Qualcomm, cheerfully relying on U.S. technology and scrapping their plans for innovation."
The above quote belies a complete misunderstanding of the current situation in China. All of Huawei's main competitors (Xiaomi, Oppo/Vivo/OnePlus, Lenovo, ZTE, etc.) procure most of their chips from US suppliers, particularly Qualcomm. That's because they've wanted to and they have to: not only is Qualcomm best in class, its main alternative (MediaTek) isn't a strong competitor at the high end. Those firms will all continue on procuring from Qualcomm (and Intel and Broadcom and Qorvo, and Skyworks, etc., etc.) if they are allowed. But unlike a year ago, if the US government decides to export control those Chinese firms too, they have a credible domestically-sourced alternative: Huawei's chips fabbed at SMIC. Based on rumors trickling out of the China tech scene, the same may soon be true for HPC needs and Nvidia. The closing of these gaps doesn't get China all the way to the cutting edge, but it buys it time. We shouldn't act like they aren't in parallel frantically pursuing a domestic EUV capability - various research and patent filings are giving glimpses of the movement on that front.
Yes, Beijing's march for self-sufficiency would have continued in either case. But Washington's moves against Huawei inadvertently (though predictably) pumped rocket fuel into that effort in a way that is already having significant unintended effects. Hand-waving away any potential for a modification of US policy to re-calibrate private sector incentives in China is a mistake. Washington still has choices.
Being on the leading edge is important for cellphones, but less so for most products. I see a lot of innovation coming from Chinese engineers and companies, there are a lot of fabs compared to what we have in the west. Engineers and scientists are saying we need to be able to work with our counterparts in China and promote the ideals of fair play, cooperation and meritocracy. We do not have the infrastructure to grow ASIC engineers in the US.
Insightful article as always. Let’s take a look at the purpose of export controls on chips. The stated objective is to constrain China from using advanced chips for military deployment. Other less clearly stated objectives are to constrain China’s overall development and stifle business competition in favour of American companies. As the objectives are conflated under “national security” it can be confusing. How sustainable are policies based on one or more of these objectives when the world is moving towards multi-polarity, challenging American hegemony?
If Mate 60 Pro had revealed holes in US export control, China would have not released it now. You think Chinese are stupid enough to expose its own weaknesses so US can tighten control accordingly? The release of Mate 60 means China have found a way to make advanced chips inspite of US export control.
ASML is already at work on building a 2nm machine. In the R&D Lab it’s working toward 1nm, and thinking in terms of picometers (pm) and femtometers (fm). Who knows what the laws of physics will allow. But the each leap to 5nm to 3nm, to 2nm is likely incrementally 10 years. Chasing each integration is exponentially difficult and can’t be achieved repeatedly running a chip through the lithography process. Who’s going to make the mirrors, understand the exact array that is different for each iteration of nanometers, & etc.? There are some things all the government subsidies can’t buy or reverse-engineer.
1- China produces more engineers than the rest of the world
2- Those engineers handicapped by various sanctions were able to produce a comparable product to engineers not handicapped by sanctions.
So, what does that mean? They have quantity and quality engineers. The more time passes the more they get closer to the critical mass that will allow them to break through some ceilings.
If you think the Chinese are not capable of creativity and innovation, then look at the next facts:
1- ASML patents have 35% Chinese names in them
2- If you look at patents outside China a lot of Chinese names are in them too.
3- If you visit research labs in the US in several fields, most of the researchers are Chinese
All what I mentioned, is never factored in conclusions drawn from China's progress in technology.
Because if they were to be included, it will contradict many myths about China.
1- Chinese government is oppressing its people
2- The communist party is incompetent
3- Chinese are low IQ and incapable of innovation and creativity
Yet, the developed tech that allowed them to:
1- Land on the dark side of the moon before everybody
2- Built a space station that will be the only one for earth in a few years
3- Landed on Mars on first shot
4- Took the lead in 5G technology
5- Developed Hyper-sonic missiles that rendered aircraft carriers obsolete
6- According to ASPI, China has dominance in 53 out of 64 critical technology.
"The most dramatic statement of confidence in China’s indigenous chip industry that I’ve seen so far comes from the blog SemiAnalysis, which predicts that Chinese chipmaker SMIC will be able to work around any and all equipment limitations."
What? That's not what we said. Please read it again.
No one seems to have commented on the time it takes to get from chip technology to chip to chip in a new product.
This is much more than 12 months so a new phone with a new chip today has nothing to do with restrictions implemented less than a year ago.
Despite some reports out there, this isn't domestication of chip making in China at all--they are still using old imported lithography machines, just in a novel way to juice out chips a few nanometers below what they are intended to make. The machines themselves are still from the Netherlands with German mirrors and full of American IP with complexity beyond the limit of human comprehension. What China is finding out now is the limits to their old equipment. It doesn't reflect a failure in sanctions regime at all, and I'd guess that their capacity to keep up with the West in chips will diverge over time. Taiwan was making 7nm chips five years ago.
Nice post. People seem to wishful-think sanctions into a magic bullet they've never been, then complain the imaginary bullet isn't lethal.
Sanctions are just a string of roadblocks. They're lousy at stopping things altogether, but pretty good at making things expensive, inconvenient, and a whole lot slower. If "expensive, inconvenient, and a whole lot slower" gives you negotiating leverage, or time to counter an opponent, that's a good case for sanctions.
If you don't really have a proposal worth negotiating, or a plan to put time on your side, sanctions aren't going to fix that for you. Sanctions exist to complement and strengthen the rest of your negotiating strategy or play-for-time strategy. They're not a strategy by themselves.
Personally I think our China strategy makes sense. Sanctions buy us more time for all our other tools to work. It's up to us to use that time well.
Sometimes, manufacturing and commitment outweigh a technology edge. Russia makes more artillery shells than the USA and Europe combined; it's even sending more drones to Ukraine than the west, despite drones being high-tech compared to explosive-filled 155mm cones of steel.
So even Russia, never mind China, is managing their quality gap with a quantity advantage.
Sanctions will buy the USA some time. But our only lasting deterrent will be if America relearns how to be an "arsenal of democracy." Billion-dollar systems built one at a time are a sign of vulnerability, not strength; the USA that'll deter China is a USA that can build military equipment not just with quality, but in quantity.
How about an update on TSMC's Arizona venture? I hear it's not going great.
The world is truly mad. What does a Chinese company producing a consumer phone have anything to do with America? So killing this innovative business seems make you all happier?
Excellent. Thanks
To me, this is a reassuring development. Not understanding the technology, I had thought that the export controls was more akin to a commodity cutoff than an important strategic move to slow down China's progress. If China believes it can overcome the export controls in time, then they are less likely to take any drastic actions. Thanks for clarifying!
robertsdavidn.substack.com.about
It's funny that in China, ignorant people like Noah would be seen as products of the bureaucratic system, while in the United States, such people are seen as experts and think tanks.
This post makes some of the same mistakes that many other commentators have on this topic (mistakes that are common on both sides of the issue, btw). Two things can be true: 1/ The 9000s isn't proof that US export controls have no effect; 2/ Huawei/SMIC have made a genuine capability breakthrough on a far faster timeline than was predicted just 2 or so years ago.
When Huawei was first added to the Entity List in 2019 and then restrictions were tightened further in 2020 to preclude its use of TSMC to fab its in-house HiSilicon chips, the state of the art was 7nm (with 5nm in sight). Lots of observation at the time pegged China as being "roughly a decade" away from being able to reach best in class. And yet, 3-4 years later, they've now reached what was (at that time) the best in class.
This has taken them less than half the time many predicted and demonstrates not so much the power of Beijing's industrial policy, but the power of what can be achieved when the incentives of China's leading private sector players (e.g., Huawei) are forcibly aligned with the Chinese state's due to US export controls. Huawei was happy to continue procuring from US suppliers (it was buying ~$20B a year!). Yes, it was developing in-house alternatives for many components as part of its longstanding strategy of vertical integration, but it saw its central role in global tech supply chains as something that was desirable, not regrettable. Until the US government intervened.
The US has now given Huawei no choice but to throw its entire self into developing credible alternatives to US and western suppliers and it's bringing its army of 100,000+ engineers to bear on the problem. This was something Beijing's industrial policy never could have accomplished on its own.
Although it's true that the cat is, to some degree out of the bag here, it would be a mistake to claim that US has no policy choice at this point and must continue on the export control course. One claim in the piece is particularly outlandish:
"It’s hilariously unrealistic to think that if the Biden administration dropped all export controls tomorrow, Chinese companies would just go right back to buying their chips from Qualcomm, cheerfully relying on U.S. technology and scrapping their plans for innovation."
The above quote belies a complete misunderstanding of the current situation in China. All of Huawei's main competitors (Xiaomi, Oppo/Vivo/OnePlus, Lenovo, ZTE, etc.) procure most of their chips from US suppliers, particularly Qualcomm. That's because they've wanted to and they have to: not only is Qualcomm best in class, its main alternative (MediaTek) isn't a strong competitor at the high end. Those firms will all continue on procuring from Qualcomm (and Intel and Broadcom and Qorvo, and Skyworks, etc., etc.) if they are allowed. But unlike a year ago, if the US government decides to export control those Chinese firms too, they have a credible domestically-sourced alternative: Huawei's chips fabbed at SMIC. Based on rumors trickling out of the China tech scene, the same may soon be true for HPC needs and Nvidia. The closing of these gaps doesn't get China all the way to the cutting edge, but it buys it time. We shouldn't act like they aren't in parallel frantically pursuing a domestic EUV capability - various research and patent filings are giving glimpses of the movement on that front.
Yes, Beijing's march for self-sufficiency would have continued in either case. But Washington's moves against Huawei inadvertently (though predictably) pumped rocket fuel into that effort in a way that is already having significant unintended effects. Hand-waving away any potential for a modification of US policy to re-calibrate private sector incentives in China is a mistake. Washington still has choices.
Being on the leading edge is important for cellphones, but less so for most products. I see a lot of innovation coming from Chinese engineers and companies, there are a lot of fabs compared to what we have in the west. Engineers and scientists are saying we need to be able to work with our counterparts in China and promote the ideals of fair play, cooperation and meritocracy. We do not have the infrastructure to grow ASIC engineers in the US.
Insightful article as always. Let’s take a look at the purpose of export controls on chips. The stated objective is to constrain China from using advanced chips for military deployment. Other less clearly stated objectives are to constrain China’s overall development and stifle business competition in favour of American companies. As the objectives are conflated under “national security” it can be confusing. How sustainable are policies based on one or more of these objectives when the world is moving towards multi-polarity, challenging American hegemony?
If Mate 60 Pro had revealed holes in US export control, China would have not released it now. You think Chinese are stupid enough to expose its own weaknesses so US can tighten control accordingly? The release of Mate 60 means China have found a way to make advanced chips inspite of US export control.
ASML is already at work on building a 2nm machine. In the R&D Lab it’s working toward 1nm, and thinking in terms of picometers (pm) and femtometers (fm). Who knows what the laws of physics will allow. But the each leap to 5nm to 3nm, to 2nm is likely incrementally 10 years. Chasing each integration is exponentially difficult and can’t be achieved repeatedly running a chip through the lithography process. Who’s going to make the mirrors, understand the exact array that is different for each iteration of nanometers, & etc.? There are some things all the government subsidies can’t buy or reverse-engineer.
This is what we know so far:
1- China produces more engineers than the rest of the world
2- Those engineers handicapped by various sanctions were able to produce a comparable product to engineers not handicapped by sanctions.
So, what does that mean? They have quantity and quality engineers. The more time passes the more they get closer to the critical mass that will allow them to break through some ceilings.
If you think the Chinese are not capable of creativity and innovation, then look at the next facts:
1- ASML patents have 35% Chinese names in them
2- If you look at patents outside China a lot of Chinese names are in them too.
3- If you visit research labs in the US in several fields, most of the researchers are Chinese
All what I mentioned, is never factored in conclusions drawn from China's progress in technology.
Because if they were to be included, it will contradict many myths about China.
1- Chinese government is oppressing its people
2- The communist party is incompetent
3- Chinese are low IQ and incapable of innovation and creativity
Yet, the developed tech that allowed them to:
1- Land on the dark side of the moon before everybody
2- Built a space station that will be the only one for earth in a few years
3- Landed on Mars on first shot
4- Took the lead in 5G technology
5- Developed Hyper-sonic missiles that rendered aircraft carriers obsolete
6- According to ASPI, China has dominance in 53 out of 64 critical technology.
"The most dramatic statement of confidence in China’s indigenous chip industry that I’ve seen so far comes from the blog SemiAnalysis, which predicts that Chinese chipmaker SMIC will be able to work around any and all equipment limitations."
What? That's not what we said. Please read it again.