Chinese semiconductor industry

hvpc · Mar 17, 2022

AndrewS said:
That sounds about right to me.

Then the next question is how many 28nm fabs they will start mass-producing every year.

All 28nm mass production investment up to 2025 go through foreign WFE supplier at the moment. All orders have mostly been communicated with suppliers including company i work for. leadtime for asml tool is longer than my company's, but what i know is SMIC is aggressively expanding 28nm fabs.

SMIC announced a 100K wafer per month fab in Beijing next to their current fabs. There's another SMIC 100K wafer per month fab in Shanghai that's being built. both should come online by 2024/2025. SMIC is also expanding their fab in Shanghai for 14nm and below. Current Shanghai fab will have up to 30ishK wafer per month of 14nm and below....they will add another 30K by 2025. All the expansion above will be with foreign WFEs. Then there's SMIC's Schenzen fab, this will also come up before 2025. From what we (WFE supplier) hear and based on ICRD timing, my guess is domestic WFE will not fully replace foreign WFE before 2025. Best guess....if everything goes according to plan, new fabs projects after 2025 is earliest we will see an all domestic equipment 28nm or below fab.

hvpc · Mar 17, 2022

Zichan said:
I am just as skeptical.

In the news yesterday, Hikvision is buying a second hand ASML's AT:850C system (248nm KrF laser), because no domestic substitute is available:

Please, Log in or Register to view URLs content!

Since early last year I was reading on this forum that a domestic DUV immersion system capable of 28nm is just around the corner (I think it was supposed to already ship by now?), but here we have a Chinese company buying 20 year old technology initially used for 110nm production and later below that. Yet, we are to believe that EUV is just 3 years away?

You hit the nail right on the head.

Like I said, SMEE doesn't have a system that's good enough for FEOL. This Hikvision purchase back-up what I said...and pretty much contradicts all those that believe China domestic EUV will be ready by 2025.

Systems that I said SMEE will release this year will be beta systems at best. You can't expect SMEE to pump out a production worthy system on its first try.

I had worked with this system while I was with the "M" memory company. This system has similar imaging capability as the latest asml KrF system. It's just much slower with less accurate overlay performance. No current SMEE tool can match this...and THAT is a fact.

hvpc · Mar 17, 2022

hvpc said:
All 28nm mass production investment up to 2025 go through foreign WFE supplier at the moment. All orders have mostly been communicated with suppliers including company i work for. leadtime for asml tool is longer than my company's, but what i know is SMIC is aggressively expanding 28nm fabs.

SMIC announced a 100K wafer per month fab in Beijing next to their current fabs. There's another SMIC 100K wafer per month fab in Shanghai that's being built. both should come online by 2024/2025. SMIC is also expanding their fab in Shanghai for 14nm and below. Current Shanghai fab will have up to 30ishK wafer per month of 14nm and below....they will add another 30K by 2025. All the expansion above will be with foreign WFEs. Then there's SMIC's Schenzen fab, this will also come up before 2025. From what we (WFE supplier) hear and based on ICRD timing, my guess is domestic WFE will not fully replace foreign WFE before 2025. Best guess....if everything goes according to plan, new fabs projects after 2025 is earliest we will see an all domestic equipment 28nm or below fab.

I forgot to mention, I have not heard Huahong group's announce any significant 28nm expansion beyond what they had added recently. Most of 28nm and below capacity expansion is by SMIC & tsmc Nanjing. I could be wrong. I'll check, if I'm wrong, I'll let you know.

56860 · Mar 17, 2022

hvpc said:
You hit the nail right on the head.

Like I said, SMEE doesn't have a system that's good enough for FEOL. This Hikvision purchase back-up what I said...and pretty much contradicts all those that believe China domestic EUV will be ready by 2025.

Systems that I said SMEE will release this year will be beta systems at best. You can't expect SMEE to pump out a production worthy system on its first try.

I had worked with this system while I was with the "M" memory company. This system has similar imaging capability as the latest asml KrF system. It's just much slower with less accurate overlay performance. No current SMEE tool can match this...and THAT is a fact.

So best case scenario is fully indigenous supply chain at 28nm by 2025, and 14nm by 2027?

hvpc · Mar 17, 2022

dengyeye said:
So best case scenario is fully indigenous supply chain at 28nm by 2025, and 14nm by 2027?

yes. that would be the most realistic timeline for someone like SMEE. Unless their engineers are many times smarter and more efficient than asml, nikon, and canon's engineers. if SMEE/ICRD can pull off this timeline I think they deserve big kudos....and this would freak out asml and canon (cuz canon's FE iline & KrF system have no significant adoption at 28nm logic either...if SMEE could pull this off they would've already jump over Canon...and Canon has come a long way on iLine & KrF..they had dominated iLine/KrF in the 90's so they have 30yrs of experience and still could not get into the 28nm and below logic space).

there's limitation on the cycle time of each testing/feedback/implement new feature iteration. it's simply not possible to get things completely right on the first iteration.

Let's look at only iLine system: Going from SMEE's BEOL system with resolution >1um and overlay requirement >100nm to FEOL level iline system that would require resolution of 330nm and overlay in the 30nm to 50nm range is simply not trivial. If it's that easy, SMEE would already be in this space....they had the SSB600/10 FEOL iLine stepper on their website for years now. If this system could really meet FE 28nm iline spec in real situation (and not just on paper), then there would already be system in production.

56860 · Mar 17, 2022

hvpc said:
yes. that would be the most realistic timeline for someone like SMEE. Unless their engineers are many times smarter and more efficient than asml, nikon, and canon's engineers. if SMEE/ICRD can pull off this timeline I think they deserve big kudos....and this would freak out asml and canon (cuz canon's FE iline & KrF system have no significant adoption at 28nm logic either...if SMEE could pull this off they would've already jump over Canon...and Canon has come a long way on iLine & KrF..they had dominated iLine/KrF in the 90's so they have 30yrs of experience and still could not get into the 28nm and below logic space).

there's limitation on the cycle time of each testing/feedback/implement new feature iteration. it's simply not possible to get things completely right on the first iteration.

Let's look at only iLine system: Going from SMEE's BEOL system with resolution >1um and overlay requirement >100nm to FEOL level iline system that would require resolution of 330nm and overlay in the 30nm to 50nm range is simply not trivial. If it's that easy, SMEE would already be in this space....they had the SSB600/10 FEOL iLine stepper on their website for years now. If this system could really meet FE 28nm iline spec in real situation (and not just on paper), then there would already be system in production.

Fully indigenous 28nm in 3 years sounds fantastic honestly.

Basically in just 3 years US is going to lose A LOT of sanctioning power.

And from there, it's only up and up.

China's exponential progress on the second most complex tech field, jet engines, has shown us that she starts slow and accelerates over time.

Yeah, WS-10 took a while. But once it was unlocked, improvements across the entire supply chain of turbofans, turboprops, turboshafts, piston engines, gas turbines, etc came rapidly. By the time WS-15 is in production, I suspect China will be 100% in sprint mode in this tech field, quickly closing on the US.

There's no reason to suspect any different for semiconductors. In fact, China might hit breakneck pace even faster because they are definitely working on EUV and DUV simultaneously.

All in all, what Japan/Nikon did in 30 years, China will achieve in a fraction of that time. Partly because catch-up growth is faster, but mainly because no country on earth can match China's depth of human capital + financial resources + laser focused leadership and vision. Once she starts running, it's game over.

hvpc · Mar 17, 2022

dengyeye said:
Fully indigenous 28nm in 3 years sounds fantastic honestly.

Basically in just 3 years US is going to lose A LOT of sanctioning power.

And from there, it's only up and up.

China's exponential progress on the second most complex tech field, jet engines, has shown us that she starts slow and accelerates over time.

Yeah, WS-10 took a while. But once it was unlocked, improvements across the entire supply chain of turbofans, turboprops, turboshafts, piston engines, gas turbines, etc came rapidly. By the time WS-15 is in production, I suspect China will be 100% in sprint mode in this tech field, quickly closing on the US.

There's no reason to suspect any different for semiconductors. In fact, China might hit breakneck pace even faster because they are definitely working on EUV and DUV simultaneously.

All in all, what Japan/Nikon did in 30 years, China will achieve in a fraction of that time. Partly because catch-up growth is faster, but mainly because no country on earth can match China's depth of human capital + financial resources + laser focused leadership and vision. Once she starts running, it's game over.

Indeed. I think if China could pull this off in 3-yr it's already very impressive.

U.S. sanction is only on EUV so she is not limiting China to buy tools for 28nm. So, China is just merely using 28nm as a learning experience to slowly drive down to 7nm and below. In order to build EUV system, China must first learn the subtle nuances of what a production scanner would need.

So even though many of us wish China could pull off EUV by 2025...we have to be realistic instead of hyping things up.

China is already falling behind on the latest 5year plan to have majority of chips made in China by 2025

Zichan said:
I am just as skeptical.

In the news yesterday, Hikvision is buying a second hand ASML's AT:850C system (248nm KrF laser), because no domestic substitute is available:

Please, Log in or Register to view URLs content!

Since early last year I was reading on this forum that a domestic DUV immersion system capable of 28nm is just around the corner (I think it was supposed to already ship by now?), but here we have a Chinese company buying 20 year old technology initially used for 110nm production and later below that. Yet, we are to believe that EUV is just 3 years away?

@Zichan, this HiKvision news turned out to be fake news:

Please, Log in or Register to view URLs content!

this is why i don't completely trust news articles.

but this doesn't take away from the fact that SMEE does not have a KrF FEOL system. SMEE will release a beta-FEOL KrF scanner some time this year (in 2022).

ZeEa5KPul · Mar 17, 2022

hvpc said:
But could turn this around and ask what is your credential to make you so upset and against information I shared? Why do you think your info is more accurate than mine? Is your push back merely based on your feeling? On educated analysis of data at hand? Or based on "facts" you read from some articles you find online?

My criterion is very simple: When a user posts that ICRD is performing verification work on EUV components and months later a paper comes out detailing exactly such work, that user has credibility. When that user quotes known SMEE insiders (not random "friends in the industry") about SMEE's progress, that user has credibility. So when that user claims that SMEE has delivered a 28nm ArF immersion lithography machine and is working on a 22nm one, I believe him. You, on the other hand...

I don't know if your "friends" are actually insiders with first-hand knowledge or voices in your head. Your problem is that you have zero credibility, so when what you say contradicts what people with credibility say, what you say goes in the trash can.

hvpc said:
I personally find most articles to be only partially accurate because they are written by layman and not from people with actual fab experience.

That you have actual fab experience is something yet to be established. Once again, your credibility problem.

hvpc said:
Just share info peacefully and you can decide for yourself if or how much of what i say you chose to believe or not.

Sharing information is great, spreading FUD and disinformation isn't. And I don't need you to tell me what to believe or not, I've already made clear that until you establish your credentials or some reason you should be taken seriously, you're just another rando with an opinion.

european_guy · Mar 17, 2022

hvpc said:
For those of you that doubt my expertise, you are welcome to ask me about semiconductor fab technology.

1) Technical question: what's about Gigaphoton and SMEE? I read that there was a collaboration in the past, but now it is no more clear if the newest litho machines in development rely on Gigaphoton's light source anymore. Maybe this has been also a factor in the delay of SMEE in the last years?

2) Random question: Why someone with such a huge and deep experience like you, decided to show it off so bluntly against us, poor amateurs / enthusiasts, and just in his firsts couple of posts? Was it your intended business card's presentation?

AndrewS · Mar 17, 2022

hvpc said:
All 28nm mass production investment up to 2025 go through foreign WFE supplier at the moment. All orders have mostly been communicated with suppliers including company i work for. leadtime for asml tool is longer than my company's, but what i know is SMIC is aggressively expanding 28nm fabs.

SMIC announced a 100K wafer per month fab in Beijing next to their current fabs. There's another SMIC 100K wafer per month fab in Shanghai that's being built. both should come online by 2024/2025. SMIC is also expanding their fab in Shanghai for 14nm and below. Current Shanghai fab will have up to 30ishK wafer per month of 14nm and below....they will add another 30K by 2025. All the expansion above will be with foreign WFEs. Then there's SMIC's Schenzen fab, this will also come up before 2025. From what we (WFE supplier) hear and based on ICRD timing, my guess is domestic WFE will not fully replace foreign WFE before 2025. Best guess....if everything goes according to plan, new fabs projects after 2025 is earliest we will see an all domestic equipment 28nm or below fab.

Assuming domestic 28nm equipment is available from 2025, how much wafer capacity could China build out from 2025-2035?

If we're going by 200MM wafer equivalents @ 28nm, China is currently at 3.5 Million WPM in terms of installed capacity.

My read is that this could be expanded to a minimum of 10 Million WPM by 2035 and that all this capacity could be consumed solely within China or placed in Chinese exports.

Chinese semiconductor industry

hvpc

Junior Member

hvpc

Junior Member

hvpc

Junior Member

56860

Senior Member

hvpc

Junior Member

56860

Senior Member

hvpc

Junior Member

ZeEa5KPul

Colonel

european_guy

Junior Member

AndrewS

Brigadier