I like the interview. What I loved was reading it - an old-school interview - as opposed to all those podcasts I can hardly stand listening to (as a non-native speaker) - coming with a computer generated transcript, full of ... errr. emm ... ya know ... incoherent sentences, restarts ... jeeezz. Gimme a break!!! You did. Highly appreciated!
I would be interested to know how this will impact a company like Tesla that is building and selling cars in China using advanced semiconductors and neural nets. Will they be required by CHIPS to use less advanced chips and therefore diminished self-drive capabilities than the same car in the US, or will they continue to build advanced technology in Gigafactory Shanghai thereby outcompeting BYD and other Chinese companies forced to use less advanced chips.
I think this is right: bifurcation is inevitable. I know much ado has been written and said about how it's overly simplistic to think of the world as "authoritarian" vs "democratic", and dollar hegemony isn't really going anywhere.... but still, this framework is incredibly useful for us.
I'm not personally convinced China will have an easy time replicating what ASML and others do any time soon, and I'm also not convinced that they will find a way around this, but I also wouldn't count out a centrally controlled government IF it decides to pivot to a much more open market system. I don't think that's in the cards today, but the winds can (and do) change rapidly in 2023.
Decent overview of the situation... missing a couple of major pieces...
1) Electronics Design Automation (EDA): This is the critical glue which makes the whole distributed supply chain work. This is dominated by US firms and unlikely to change. Without EDA, you cannot design even the simplest chip and a great deal of the most critical "IP" is embedded in these EDA tools. EDA also forms the backbone of the economic interactions which makes a company like TSMC function.
2) Innovation vs Follow-me cultures: There is a fundamentally different machinery one builds when one is behind (china, japan in the past, etc). In a follow-me culture, the key skills are tracking the leader in legal and illegal ways. However, it is entirely a different set of skills to be on the frontier and innovate in open-space. This is the reason one sees a fast catchup in many countries and then a flattening process.
Agree.. need both... though a modern fab melds the two in interesting ways. On the manufacturing side, keep an eye on developments in Mexico.
Also, while the hardware certainly forms the basis of computing, a great deal of the innovation is happening at higher levels of abstraction (SW, AI). There are a lot of big levers in this space. Just to give you a sense of it.. a modern AV (which is barely autonomous) consumes hundreds of watts of power...while a human brain is around a 10W. At least on order-of-magnitude better..likely two. This is a way of saying that some big breakthroughs in the area of SW/AI on the algorithms side change the dynamics on the HW/fab side pretty drastically. Today, most of SW/AI is leveraging HW computation to discover interesting capabilities.
To certain extent they did, or at least they lacked enough capability. Ultimately (and thankfully) it caught up with them. The Germans had a huge issue with believing that technological and tactical innovations could overcome strategic limitations. I fear that we we may fall into a similar trap.
We have de-industrialized because the stock market values high margins, IP, low fixed capital costs and continual focus on incremental cost savings.
This was a matter of choice (and an economic model focused on stock price) rather than capacity and raw material limitations, like Germany.
I had American friends who graduated in the 1980s and designed clean rooms and litho facilities for chip manufacturing for American firms. Even then most of the facilities they designed were being built overseas.
To some extent we are trying to push China into that German box - dependent on imports that can be limited- while China tries (successfully) to make the rest of the world dependent upon it
- a strategy Germany didn’t try. China is also using bribes and loans in the developing world (and with OECD pols - paying the Clintons, Biden’s, Kerrys) while Germany took more of the old imperialist/colonialist approach (not that the end result for the emerging country will be much different)
Interesting to see China transition from an “island” strategy to great power competition. Trump was the sea change in the US approach. Whether it came to late or not- we’ll see.
Innovation vs. Follow-me cultures... While reading this interview the old adage: "Necessity is the mother of invention..." kept running through my head. Yes, until now China could have the luxury of being a Follow-me culture regarding advanced chips, but we are creating a necessity for them to switch to being Innovators. As we have seen in other industries, China can be very creative if need be. I sense we are foolishly awakening the Dragon.
My sense is that it is not that easy. To do it well, you need some background of IP protection, a different sort of education system, and more fundamentally a cultural shift.
My sense is that they have a long way to do...especially in the context of a repressive governmental structure.
At first I thought you were describing with the U.S. needed to improve. I am betting your opinions about China are very outdated. Check out these charts as to where China stands with IP and STEM education.
I wouldn't take the IP and STEM education statistics from China too seriously. Ignoring the legal infrastructure problems with leveraging IP, the question is what the real value of these patents ?. Similarly, in STEM education, there is a factory approach which produces a lot of questionable degrees (you see these STEM mills out of India as well). You can see the innovation rate difference in the major breakthrough technologies. This is one of the reasons that youth unemployment is going through the roof.
BTW... I was the Sr VP of Strategy for a public company that employed over 300K employees in China. At least in electronics/SW/AI area, I have direct experience.
I hope you right. However, I fear you are not. All the vectors are pointing the wrong way right now (debt, demographics, political repression, world economic derisking/decoupling, loss of cost competitiveness to southeast asia and others). I fear it ends badly, but hope there is a soft landing of some sort.
Good interview. During the long stretch of years when ASML progressively developed more advanced EUV lithography machines, something else was happening that people may not know.
• 90% of ASML machines are still working. I’m fact, there is a resale market.
• unlike most companies, ASML continues to support machines with repairs and parts
• nobody has mastered supply-chain management to the extent of ASML, with more than 4,000 vendors
• these days, when you buy a high-end ASML, the service contracts are a revenue source, because ASML engineers/technicians live nearby
• in re security, ASML has an in-house espionage team of more than 100 employees
• only a few key people at ASML understand how the machine works. Employee teams build a certain portion of the machine, not knowing exactly how it integrates/functions together with other portions.
• during the pandemic, ASML engineers guided, via a Zoom call, a client through a successful repair.
• reverse-engineering a machine that requires four 747 Freightliner aircraft to ship all the parts may be a bridge too far for China
• the expertise gap will continue to grow. The 2nm machine begins use toward the end of this year at TSMC
•1nm and smaller is already in development in the R&D lab in The Netherlands.
Excellent interview, thanks! Good overview and seems like a reasonable/thoughtful scenario for the next several years .
Sounds like the US is going to be dependent upon Taiwan and S Korea (and Holland) for years to come (and if I were Taiwanese this is exactly what I would want), unless Intel or a new entrant is able to catch up in leading edge fabrication.
Hopefully China shoots itself in the foot with its “buy China” policies. We’ll see. From a national and industrial security point of view, not sure his idea of importing thousands of mainland Chinese engineers is a great idea, however. The US needs to invest in training (along with immigration).
This is a very thoughtful and intriguing piece but it is missing one key "element". I note element as the article does not address the fact that China basically controls all the elements (critical metals and rare earths). The tech industry at large has not addressed this issue at all nor have North American governments. Despite both the United States and Canada having much of these industry and the capital markets have not invested or looked at this monopolistic bottleneck. Just last week China restricted the export of gallium and germanium. So build away, but underinvestment in the actual mining of the elements that make chips will be the ultimate loss.
The Gallium/Germanium controls are a licensing system rather than an outright ban. It remains to be seen how the licensing regime will work, which goes into effect August 1st.
While clearly a tit-for-tat move, the effect is not expected to be nearly as dramatic as the Western export controls.
The elements themselves are found all over. The reason there is less production in the West is mainly due to tougher environmental regulations. In essence, the issue is more one of will than anything else. On the other hand, the Western controls are much harder to overcome, requiring substantial development of new knowhow.
Fascinating stuff. The regulation point reminded me of something I heard on NPR years ago about how every US state requires hairstylists to get certified and licensed. Not exactly semiconductor-level stuff but perhaps our culture of regulation has gone a bit far.
Just wanted to know if US companies like Tesla, Apple, Google can build their own CHIP from scratch then why can't Chinese ones like Baidu, Huawei etc. Thank you.
None of those companies make their own chips. Apple designs it’s M and A series chips used in Macs and mobile devices, but they are manufactured by TSMC in Taiwan. Google designs chips for data centers like its Tensor Flow TPUs but again they are manufactured by TSMC. Tesla also designs chips but doesn’t manufacture them. The interview didn’t go into design vs manufacturing but US leads in design and design software like Synopsis and Cadence and Ansys. Siemens in Germany also had design software, and China may be trying to develop their own, but I don’t think the Biden export controls cover design software or services.
3. TSMC’s 3nm yield rate reportedly just 55%, with Apple only paying for qualified circuits
“TSMC is struggling with the efficiency of its new 3nm manufacturing yield, with the semiconductor giant currently hitting a yield rate of just 55%, far below the standard expected, according to a July 13 report in technology media outlet wccftech. The low yield rate has reportedly led Apple to only pay for qualified wafer batches instead of establishing a standard rate with TSMC. Apple occupies 90% of TSMC’s 3nm process production capacity for its A17 Bionic and M3 chips.” -- Tech Node
Am I the only one who saw the picture of Chris Miller and thought, “how old is this kid?!”
I think he's 36
I freely admit that it's incredibly It's superficial, but also: yes, this was my first reaction, too.
I like the interview. What I loved was reading it - an old-school interview - as opposed to all those podcasts I can hardly stand listening to (as a non-native speaker) - coming with a computer generated transcript, full of ... errr. emm ... ya know ... incoherent sentences, restarts ... jeeezz. Gimme a break!!! You did. Highly appreciated!
Life is too short for Podcasts, even at 2x. Good writers seem to assume they will be interesting and professional speakers, but few are.
To be fair, if I had a long commute by car podcasts might be useful. Sometimes my playlists get old
I would be interested to know how this will impact a company like Tesla that is building and selling cars in China using advanced semiconductors and neural nets. Will they be required by CHIPS to use less advanced chips and therefore diminished self-drive capabilities than the same car in the US, or will they continue to build advanced technology in Gigafactory Shanghai thereby outcompeting BYD and other Chinese companies forced to use less advanced chips.
This was fascinating, thank you!
> it takes longer to build a fab in the U.S. than in Europe
There hasn't been a SOTA fab built in Europe in over a decade so this is surprising. The last one was opened by GloFo in 2011.
I think this is right: bifurcation is inevitable. I know much ado has been written and said about how it's overly simplistic to think of the world as "authoritarian" vs "democratic", and dollar hegemony isn't really going anywhere.... but still, this framework is incredibly useful for us.
I'm not personally convinced China will have an easy time replicating what ASML and others do any time soon, and I'm also not convinced that they will find a way around this, but I also wouldn't count out a centrally controlled government IF it decides to pivot to a much more open market system. I don't think that's in the cards today, but the winds can (and do) change rapidly in 2023.
Thanks for some great food for thought!
I can tell you that whether China can buy EUV depends entirely on Biden, not on what the CCP has done.
Decent overview of the situation... missing a couple of major pieces...
1) Electronics Design Automation (EDA): This is the critical glue which makes the whole distributed supply chain work. This is dominated by US firms and unlikely to change. Without EDA, you cannot design even the simplest chip and a great deal of the most critical "IP" is embedded in these EDA tools. EDA also forms the backbone of the economic interactions which makes a company like TSMC function.
2) Innovation vs Follow-me cultures: There is a fundamentally different machinery one builds when one is behind (china, japan in the past, etc). In a follow-me culture, the key skills are tracking the leader in legal and illegal ways. However, it is entirely a different set of skills to be on the frontier and innovate in open-space. This is the reason one sees a fast catchup in many countries and then a flattening process.
Imagine if the Germans during WW2, with the best design engineering, had no industrial capacity to manufacture their designs.
Agree.. need both... though a modern fab melds the two in interesting ways. On the manufacturing side, keep an eye on developments in Mexico.
Also, while the hardware certainly forms the basis of computing, a great deal of the innovation is happening at higher levels of abstraction (SW, AI). There are a lot of big levers in this space. Just to give you a sense of it.. a modern AV (which is barely autonomous) consumes hundreds of watts of power...while a human brain is around a 10W. At least on order-of-magnitude better..likely two. This is a way of saying that some big breakthroughs in the area of SW/AI on the algorithms side change the dynamics on the HW/fab side pretty drastically. Today, most of SW/AI is leveraging HW computation to discover interesting capabilities.
Interesting, thanks
To certain extent they did, or at least they lacked enough capability. Ultimately (and thankfully) it caught up with them. The Germans had a huge issue with believing that technological and tactical innovations could overcome strategic limitations. I fear that we we may fall into a similar trap.
We have de-industrialized because the stock market values high margins, IP, low fixed capital costs and continual focus on incremental cost savings.
This was a matter of choice (and an economic model focused on stock price) rather than capacity and raw material limitations, like Germany.
I had American friends who graduated in the 1980s and designed clean rooms and litho facilities for chip manufacturing for American firms. Even then most of the facilities they designed were being built overseas.
To some extent we are trying to push China into that German box - dependent on imports that can be limited- while China tries (successfully) to make the rest of the world dependent upon it
- a strategy Germany didn’t try. China is also using bribes and loans in the developing world (and with OECD pols - paying the Clintons, Biden’s, Kerrys) while Germany took more of the old imperialist/colonialist approach (not that the end result for the emerging country will be much different)
Interesting to see China transition from an “island” strategy to great power competition. Trump was the sea change in the US approach. Whether it came to late or not- we’ll see.
Innovation vs. Follow-me cultures... While reading this interview the old adage: "Necessity is the mother of invention..." kept running through my head. Yes, until now China could have the luxury of being a Follow-me culture regarding advanced chips, but we are creating a necessity for them to switch to being Innovators. As we have seen in other industries, China can be very creative if need be. I sense we are foolishly awakening the Dragon.
My sense is that it is not that easy. To do it well, you need some background of IP protection, a different sort of education system, and more fundamentally a cultural shift.
My sense is that they have a long way to do...especially in the context of a repressive governmental structure.
At first I thought you were describing with the U.S. needed to improve. I am betting your opinions about China are very outdated. Check out these charts as to where China stands with IP and STEM education.
https://www.wipo.int/en/ipfactsandfigures/patents
I wouldn't take the IP and STEM education statistics from China too seriously. Ignoring the legal infrastructure problems with leveraging IP, the question is what the real value of these patents ?. Similarly, in STEM education, there is a factory approach which produces a lot of questionable degrees (you see these STEM mills out of India as well). You can see the innovation rate difference in the major breakthrough technologies. This is one of the reasons that youth unemployment is going through the roof.
BTW... I was the Sr VP of Strategy for a public company that employed over 300K employees in China. At least in electronics/SW/AI area, I have direct experience.
I bow to your experience, but also acknowledge my deep intuition. Modern China is taking baby steps now, but will likely make giant strides later.
I hope you right. However, I fear you are not. All the vectors are pointing the wrong way right now (debt, demographics, political repression, world economic derisking/decoupling, loss of cost competitiveness to southeast asia and others). I fear it ends badly, but hope there is a soft landing of some sort.
Thanks for sharing this article. Quite interesting ! Let's see how future unfolds for the industry and what role India really plays down the line
Good interview. During the long stretch of years when ASML progressively developed more advanced EUV lithography machines, something else was happening that people may not know.
• 90% of ASML machines are still working. I’m fact, there is a resale market.
• unlike most companies, ASML continues to support machines with repairs and parts
• nobody has mastered supply-chain management to the extent of ASML, with more than 4,000 vendors
• these days, when you buy a high-end ASML, the service contracts are a revenue source, because ASML engineers/technicians live nearby
• in re security, ASML has an in-house espionage team of more than 100 employees
• only a few key people at ASML understand how the machine works. Employee teams build a certain portion of the machine, not knowing exactly how it integrates/functions together with other portions.
• during the pandemic, ASML engineers guided, via a Zoom call, a client through a successful repair.
• reverse-engineering a machine that requires four 747 Freightliner aircraft to ship all the parts may be a bridge too far for China
• the expertise gap will continue to grow. The 2nm machine begins use toward the end of this year at TSMC
•1nm and smaller is already in development in the R&D lab in The Netherlands.
Excellent interview, thanks! Good overview and seems like a reasonable/thoughtful scenario for the next several years .
Sounds like the US is going to be dependent upon Taiwan and S Korea (and Holland) for years to come (and if I were Taiwanese this is exactly what I would want), unless Intel or a new entrant is able to catch up in leading edge fabrication.
Hopefully China shoots itself in the foot with its “buy China” policies. We’ll see. From a national and industrial security point of view, not sure his idea of importing thousands of mainland Chinese engineers is a great idea, however. The US needs to invest in training (along with immigration).
This is a very thoughtful and intriguing piece but it is missing one key "element". I note element as the article does not address the fact that China basically controls all the elements (critical metals and rare earths). The tech industry at large has not addressed this issue at all nor have North American governments. Despite both the United States and Canada having much of these industry and the capital markets have not invested or looked at this monopolistic bottleneck. Just last week China restricted the export of gallium and germanium. So build away, but underinvestment in the actual mining of the elements that make chips will be the ultimate loss.
The Gallium/Germanium controls are a licensing system rather than an outright ban. It remains to be seen how the licensing regime will work, which goes into effect August 1st.
While clearly a tit-for-tat move, the effect is not expected to be nearly as dramatic as the Western export controls.
The elements themselves are found all over. The reason there is less production in the West is mainly due to tougher environmental regulations. In essence, the issue is more one of will than anything else. On the other hand, the Western controls are much harder to overcome, requiring substantial development of new knowhow.
Fascinating stuff. The regulation point reminded me of something I heard on NPR years ago about how every US state requires hairstylists to get certified and licensed. Not exactly semiconductor-level stuff but perhaps our culture of regulation has gone a bit far.
Such a fantastic interview, thanks Noah!
Just wanted to know if US companies like Tesla, Apple, Google can build their own CHIP from scratch then why can't Chinese ones like Baidu, Huawei etc. Thank you.
None of those companies make their own chips. Apple designs it’s M and A series chips used in Macs and mobile devices, but they are manufactured by TSMC in Taiwan. Google designs chips for data centers like its Tensor Flow TPUs but again they are manufactured by TSMC. Tesla also designs chips but doesn’t manufacture them. The interview didn’t go into design vs manufacturing but US leads in design and design software like Synopsis and Cadence and Ansys. Siemens in Germany also had design software, and China may be trying to develop their own, but I don’t think the Biden export controls cover design software or services.
China has Loongson（龙芯）, but it still has a long way to go in terms of both performance and market share.
I enjoyed reading through this interview. Thanks for doing it and leaving it as free to read.
3. TSMC’s 3nm yield rate reportedly just 55%, with Apple only paying for qualified circuits
“TSMC is struggling with the efficiency of its new 3nm manufacturing yield, with the semiconductor giant currently hitting a yield rate of just 55%, far below the standard expected, according to a July 13 report in technology media outlet wccftech. The low yield rate has reportedly led Apple to only pay for qualified wafer batches instead of establishing a standard rate with TSMC. Apple occupies 90% of TSMC’s 3nm process production capacity for its A17 Bionic and M3 chips.” -- Tech Node