TechanaLye assesses China as 3 years behind TSMC

[KevinK]

SMIC doubled the size of its FinFET production facilities. We are supposed to believe there are chip supply issues. I doubt it.

Doubling the facility size doesn’t fill it with embargoed equipment, or deliver spare parts. There have been a number of reports from various sources highlighting SMIC purported travails - quad patterning driving far more litho steps and equipment wear and tear, plus poor yields. But what’s especially telling are the facts on the ground - far smaller production runs for Kirin smartphone chips and 910B AI chips than expected. Again, we’ll see the volumes for the Mate 70 when it is finally launched, if the numbers are anything like the earlier phones using Kirin variants, 10M units or less, we’ll know that they are over a barrel.

“Huawei and Semiconductor Manufacturing International realized they did not have enough parts to increase manufacturing of the chips. This could make it difficult for Huawei to continue even a smaller-scale manufacturing and to make successor chips to the 910B, as it had planned”

“In April, it was reported that due to rising demand for AI chips and production constraints, Huawei was focusing on AI and scaling back production of its flagship Mate 60 phones.”

https://www.theinformation.com/articles/a-new-u-s-crackdown-is-crippling-chinas-best-hope-to-rival-nvidia

 

[samwilde]

Doubling the facility size doesn’t fill it with embargoed equipment, or deliver spare parts.

SMIC could have redeployed the immersion lithography machines they already had at their other facilities to make the simpler layers. And used the leading edge machines at their FinFET fab to make the harder layers. The import ban was only applied after the fab expansion shell was complete as well. Who knows how much equipment they got in before the ban.

Then you have the tool vendors moving sales to China ahead of everyone else in deliveries to ensure they could get as much money as possible.

Notice how Singapore is exporting more chip making tools to China than the US with Malaysia right next after the US. That is another loophole. US tool vendors like Applied Materials had fabs in Singapore and Malaysia and were shipping the tools to China anyway, ban or no ban, that landed them into trouble with the US government recently.

Bloomberg - Are you a robot?

There was also a massive gap in sanctions where South Korean third parties were reselling machines to China after the ban from the tool vendors. This lasted for quite a while until they clamped down on it. Quite likely South Korean DRAM vendors were dumping immersion DUV equipment in expectation of a move to EUV. Those machines could also have ended up at SMIC.

I would expect SMIC to have those machines and that fab capacity. They probably are in the process of ramping the fab expansion up. As for the spare parts there are third parties which do tool maintenance.

There have been a number of reports from various sources highlighting SMIC purported travails - quad patterning driving far more litho steps and equipment wear and tear, plus poor yields.

Hearsay and people making things up. TechInsights imaged the SMIC made chips and in their opinion the way they look does not imply low yields at all.

But what’s especially telling are the facts on the ground - far smaller production runs for Kirin smartphone chips and 910B AI chips than expected.

Huawei sold tens of millions of smartphones. Also it is not like SMIC only makes chips for Huawei. For example they make CPUs for Loongson and Phytium with the FinFET process. The Chinese government put a law into place where all government computers must use Chinese chips. They are replacing all their computers with Chinese made ones. Imagine every single government clerk there getting a Chinese PC.
All of those hardware sales are invisible. But they sure show up on Intel and AMD's balance sheets as lost sales in China.

Again, we’ll see the volumes for the Mate 70 when it is finally launched, if the numbers are anything like the earlier phones using Kirin variants, 10M units or less, we’ll know that they are over a barrel.

Since when is 10M sales of any semiconductor chip a bad outcome?

“In April, it was reported that due to rising demand for AI chips and production constraints, Huawei was focusing on AI and scaling back production of its flagship Mate 60 phones.”

It couldn't possibly be that they are scaling back production of the Mate 60 phones because it is a one year old product at this point and they already have enough stock to satisfy existing demand. Right?

 

[jim_22222]

SMIC could have redeployed the immersion lithography machines they already had at their other facilities to make the simpler layers. And used the leading edge machines at their FinFET fab to make the harder layers. The import ban was only applied after the fab expansion shell was complete as well. Who knows how much equipment they got in before the ban.

Then you have the tool vendors moving sales to China ahead of everyone else in deliveries to ensure they could get as much money as possible.

Notice how Singapore is exporting more chip making tools to China than the US with Malaysia right next after the US. That is another loophole. US tool vendors like Applied Materials had fabs in Singapore and Malaysia and were shipping the tools to China anyway, ban or no ban, that landed them into trouble with the US government recently.

Bloomberg - Are you a robot?

There was also a massive gap in sanctions where South Korean third parties were reselling machines to China after the ban from the tool vendors. This lasted for quite a while until they clamped down on it. Quite likely South Korean DRAM vendors were dumping immersion DUV equipment in expectation of a move to EUV. Those machines could also have ended up at SMIC.

I would expect SMIC to have those machines and that fab capacity. They probably are in the process of ramping the fab expansion up. As for the spare parts there are third parties which do tool maintenance.


Hearsay and people making things up. TechInsights imaged the SMIC made chips and in their opinion the way they look does not imply low yields at all.


Huawei sold tens of millions of smartphones. Also it is not like SMIC only makes chips for Huawei. For example they make CPUs for Loongson and Phytium with the FinFET process. The Chinese government put a law into place where all government computers must use Chinese chips. They are replacing all their computers with Chinese made ones. Imagine every single government clerk there getting a Chinese PC.
All of those hardware sales are invisible. But they sure show up on Intel and AMD's balance sheets as lost sales in China.


Since when is 10M sales of any semiconductor chip a bad outcome?


It couldn't possibly be that they are scaling back production of the Mate 60 phones because it is a one year old product at this point and they already have enough stock to satisfy existing demand. Right?

I would say euv break through would take two to three years for next check point. no need to check this every quarter etc. if no big news two or three years later some thing wrong with their system.

 

[mattyenright]

I would say euv break through would take two to three years for next check point. no need to check this every quarter etc. if no big news two or three years later some thing wrong with their system.

EUV breakthrough? What are you referring too.

 

[jim_22222]

EUV breakthrough? What are you referring too.

...
long story, if some one want consultation on this subject can message me.

 

[KevinK]

Since when is 10M sales of any semiconductor chip a bad outcome?

When you are leading supplier like Apple or Samsung, where new phones sell 70M units and 30M units respectively in the first months after launch.

It couldn't possibly be that they are scaling back production of the Mate 60 phones because it is a one year old product at this point and they already have enough stock to satisfy existing demand. Right?

Tapped out at only 10M units ?

 

[hasty2333]

Thanks to the lively punch-up in this thread, it really hammers home how semiconductors are just as political as they are technical

 

[Fred Chen]

SMIC could have redeployed the immersion lithography machines they already had at their other facilities to make the simpler layers. And used the leading edge machines at their FinFET fab to make the harder layers. The import ban was only applied after the fab expansion shell was complete as well. Who knows how much equipment they got in before the ban.

Then you have the tool vendors moving sales to China ahead of everyone else in deliveries to ensure they could get as much money as possible.

Notice how Singapore is exporting more chip making tools to China than the US with Malaysia right next after the US. That is another loophole. US tool vendors like Applied Materials had fabs in Singapore and Malaysia and were shipping the tools to China anyway, ban or no ban, that landed them into trouble with the US government recently.

Bloomberg - Are you a robot?

There was also a massive gap in sanctions where South Korean third parties were reselling machines to China after the ban from the tool vendors. This lasted for quite a while until they clamped down on it. Quite likely South Korean DRAM vendors were dumping immersion DUV equipment in expectation of a move to EUV. Those machines could also have ended up at SMIC.

I would expect SMIC to have those machines and that fab capacity. They probably are in the process of ramping the fab expansion up. As for the spare parts there are third parties which do tool maintenance.


Hearsay and people making things up. TechInsights imaged the SMIC made chips and in their opinion the way they look does not imply low yields at all.


Huawei sold tens of millions of smartphones. Also it is not like SMIC only makes chips for Huawei. For example they make CPUs for Loongson and Phytium with the FinFET process. The Chinese government put a law into place where all government computers must use Chinese chips. They are replacing all their computers with Chinese made ones. Imagine every single government clerk there getting a Chinese PC.
All of those hardware sales are invisible. But they sure show up on Intel and AMD's balance sheets as lost sales in China.


Since when is 10M sales of any semiconductor chip a bad outcome?


It couldn't possibly be that they are scaling back production of the Mate 60 phones because it is a one year old product at this point and they already have enough stock to satisfy existing demand. Right?

The domestic Kirin chips released up to now have been 7nm-class ("N+2"), so no quadruple patterning or specifically pitch quartering would have been needed. The process would have been running over a year now. But the volumes might be limited by how the tools are allocated.

 

[Fred Chen]

I would say euv break through would take two to three years for next check point. no need to check this every quarter etc. if no big news two or three years later some thing wrong with their system.

However, their DUV system announcements (or lack thereof in some cases) are already not so impressive.

 

[KevinK]

Thanks to the lively punch-up in this thread, it really hammers home how semiconductors are just as political as they are technical

It’s not often SMIC makes the New York Times

The Chinese Chipmaker at the Heart of the U.S.-China Tech War​

Efforts by the Beijing-backed Semiconductor Manufacturing International Corporation, or SMIC, to break through innovation barriers have landed it in a geopolitical tech battle.

Dylan Patel, the chief analyst at SemiAnalysis, a research firm, estimated that SMIC could make 1.2 million A.I. chips for Huawei next year, double this year — far fewer than what China needs or what American chip designers make, but indicating an upward trajectory.

https://www.nytimes.com/2024/09/16/technology/smic-china-us-trade-war.html?unlocked_article_code=1.ME4.FA3k.DTIcsBDLIDYK&smid=nytcore-ios-share&referringSource=articleShare&sgrp=c-cb&ngrp=mnp

 

[jim_22222]

However, their DUV system announcements (or lack thereof in some cases) are already not so impressive.

You need to remember that they abandoned this program and picked it up again after 2018. They were in a quasi-honeymoon mode with the West during the Obama administration, and there was little incentive to develop those tools.

提示信息

It takes time to build up the momentum and the right culture to cultivate this type of business. The toughest part is the cultural transition, from a workaholic 996/711 mode to a more healthy balanced mode, which is tough.

TSMC's success is built on attracting top executives and engineers from companies like TI. If mainland China learns this strategy, they will grow quickly.

 

[mattyenright]

You need to remember that they abandoned this program and picked it up again after 2018. They were in a quasi-honeymoon mode with the West during the Obama administration, and there was little incentive to develop those tools.

提示信息

It takes time to build up the momentum and the right culture to cultivate this type of business. The toughest part is the cultural transition, from a workaholic 996/711 mode to a more healthy balanced mode, which is tough.

TSMC's success is built on attracting top executives and engineers from companies like TI. If mainland China learns this strategy, they will grow quickly.

I apologise for the blunt response, but I don't think that is fully accurate. Since I can remember various plans and initiatives for complete technological independence in both the semiconductor equipment and semiconductor manufacturing disciplines that we're directly embedded into various national state initiatives and outlines such as 02 Special Project: (2009-2011), Made-in-China-2025 (2015), National Semiconductor Industry Development Guidelines (2014), and the China's 13th Five-Year Plan for S&T Innovation (2015).

These aforementioned initiatives contained specific aims such as attaining 40% self-sufficiency in China’s total IC consumption by 2020, and 70% by 2025. Construction of 14nm etching equipment and 28nm capable lithography machines by 2020-2025. Along with building a completely independent private and public ecosystem that was international competitive by 2030.

 

[jim_22222]

I apologise for the blunt response, but I don't think that is fully accurate. Since I can remember various plans and initiatives for complete technological independence in both the semiconductor equipment and semiconductor manufacturing disciplines that we're directly embedded into various national state initiatives and outlines such as 02 Special Project: (2009-2011), Made-in-China-2025 (2015), National Semiconductor Industry Development Guidelines (2014), and the China's 13th Five-Year Plan for S&T Innovation (2015).

These aforementioned initiatives contained specific aims such as attaining 40% self-sufficiency in China’s total IC consumption by 2020, and 70% by 2025. Construction of 14nm etching equipment and 28nm capable lithography machines by 2020-2025. Along with building a completely independent private and public ecosystem that was international competitive by 2030.

those plan mostly stayed on paper.
with very limited funding.
if you have the number of funding they put into it you will know it is mostly some concept proving efforts.
at the same time, there were another group of intellectuals playing down the risk of international supply chain and promoting collaboration with the west.

which looked promising as well. can you imagine UC Berkeley sent dozens of its top faculties to china to work on projects.

you also can't say Berkeley want to sell every thing to china just because they had that collaboration. you need to look into the details .