Chinese semiconductor thread II

[tphuang]

Fake news

The construction plan for ZETOP is documented in a detailed Word file. I contacted the Changchun government for consultation (with call recordings available for verification). They provided me with the detailed document only after I assured them I would not disclose confidential information. The screenshot above shows only the publicly available portion; I will not release any confidential sections.

There are multiple EUV project going on in China. There is no evidence one way or the other what the so called 2025Q3 rumor is true or not. Since we don't even know what that rumor is.

In fact, my contact has sent me a picture of a facility in Dongguan earlier this year that he claimed to be the EUV facility. Or at least that's where they say Huawei affiliated people are testing at.

Now, that does not mean the EUV project is ready for production. Most likely, we are still some time away. You need to have a facility before you can move machines in and test trialing a process with it. All of this takes time.

But this is certainly not something that Changchun gov't would know. This is something that people with connection in Shenzhen tech space would know.

Clearly by this point, SiCarrier's DUVi machine is ahead of CETC and SMEE in performance. I would be shocked if Huawei affiliated EUV project would not be ahead of the project being done in Northern China.

you can hook up with your insider for EUV info?
A while back rumors of EUV coming on line 3rd quarter of 2025.

It would be good for people to stop asking these questions, since nobody knows for sure and we will get a wild goose chase every time. The opsec is very tight and beyond the scope of people on this forum. And frankly, it makes my life difficult.

In fact, you have been the source of many of initiating these goose chases. So, I would suggest that you tighten up the quality of posting a little bit or we will need to ban you from this thread.

 

[Awenumick]

View attachment 160978
The timeline for DUV lithography machines is 2025, while EUV lithography machines are scheduled for 2027 to 2030. This is the national plan, and I'm unsure if there will be any acceleration.
The national medium-to-long-term plan includes a roadmap, with EUV lithography machines slated for realization by 2030.
Moreover, due to the current tensions between China and the United States, all projects are subject to extremely high levels of secrecy. Simply knowing the timelines is already quite challenging.

What is the source of this image?

 

[jx191]

There are multiple EUV project going on in China. There is no evidence one way or the other what the so called 2025Q3 rumor is true or not. Since we don't even know what that rumor is.

In fact, my contact has sent me a picture of a facility in Dongguan earlier this year that he claimed to be the EUV facility. Or at least that's where they say Huawei affiliated people are testing at.

Now, that does not mean the EUV project is ready for production. Most likely, we are still some time away. You need to have a facility before you can move machines in and test trialing a process with it. All of this takes time.

But this is certainly not something that Changchun gov't would know. This is something that people with connection in Shenzhen tech space would know.

Clearly by this point, SiCarrier's DUVi machine is ahead of CETC and SMEE in performance. I would be shocked if Huawei affiliated EUV project would not be ahead of the project being done in Northern China.


It would be good for people to stop asking these questions, since nobody knows for sure and we will get a wild goose chase every time. The opsec is very tight and beyond the scope of people on this forum. And frankly, it makes my life difficult.

In fact, you have been the source of many of initiating these goose chases. So, I would suggest that you tighten up the quality of posting a little bit or we will need to ban you from this thread.

I think he was referring to specifically ZETOP rather than the full EUV project. Like you said, local government wouldn't have the access to the EUV project information, but construction and activity by ZETOP sounds like it would have less opsec than the bigger parts of the project.

 

[tphuang]

I think he was referring to specifically ZETOP rather than the full EUV project. Like you said, local government wouldn't have the access to the EUV project information, but construction and activity by ZETOP sounds like it would have less opsec than the bigger parts of the project.

I'm referring to the general post here "The timeline for DUV lithography machines is 2025, while EUV lithography machines are scheduled for 2027 to 2030. This is the national plan, and I'm unsure if there will be any acceleration."

The point is nobody outside of the people testing this stuff in Shenzhen area knows how far along they are. So at best, we can say that we don't know how far they are. but clearly based on the Ai chip roadmap by Huawei today, this is not something we need to worry that much about. Because the people inside Huawei itself aren't worried.

as @latenlazy said before, EUV project has lower hurdles before going into production usage, because it's main competition is multi-patterning with Arfi scanners. As such, it just needs to work better than multi-patterning to be used in a production process. So yes, a Arfi scanner may take another year or two for SMIC and Huawei to feel comfortable with 7nm. But EUV project is not competing against ASML EUV scanners.


This particular patent was filed in November of 2024 and it's on the exposure optical system of EUV project. So it seems to me there is no particular delays in this related area. It's likely being tested out right now just like rest of the system.

What I heard a while back is that they could start risk production using EUV line late this year. however, that is likely a very optimistic take. And more importantly, risk production is not the same as HVM. It also takes months to finish each batch of chips. And they are going to be testing and fixing along.

But if you want to be optimistic, just look at the specs that Huawei posted. You need some really advanced node to produce the Serdes chip needed for the 4 TB/s interconnect they have planned for Ascend-970. And you need some seriously advanced node for the HBM they have listed for Ascend-970.

 

[jx191]

I'm referring to the general post here "The timeline for DUV lithography machines is 2025, while EUV lithography machines are scheduled for 2027 to 2030. This is the national plan, and I'm unsure if there will be any acceleration."

The point is nobody outside of the people testing this stuff in Shenzhen area knows how far along they are. So at best, we can say that we don't know how far they are. but clearly based on the Ai chip roadmap by Huawei today, this is not something we need to worry that much about. Because the people inside Huawei itself aren't worried.

as @latenlazy said before, EUV project has lower hurdles before going into production usage, because it's main competition is multi-patterning with Arfi scanners. As such, it just needs to work better than multi-patterning to be used in a production process. So yes, a Arfi scanner may take another year or two for SMIC and Huawei to feel comfortable with 7nm. But EUV project is not competing against ASML EUV scanners.

View attachment 160984
This particular patent was filed in November of 2024 and it's on the exposure optical system of EUV project. So it seems to me there is no particular delays in this related area. It's likely being tested out right now just like rest of the system.

What I heard a while back is that they could start risk production using EUV line late this year. however, that is likely a very optimistic take. And more importantly, risk production is not the same as HVM. It also takes months to finish each batch of chips. And they are going to be testing and fixing along.

But if you want to be optimistic, just look at the specs that Huawei posted. You need some really advanced node to produce the Serdes chip needed for the 4 TB/s interconnect they have planned for Ascend-970. And you need some seriously advanced node for the HBM they have listed for Ascend-970.

Thanks for the clarification. I remember @tokenanalyst pointing out that whilst the progress of EUV development cannot be confirmed, the publications of research, testing and development are good indicators of how far along they are, especially considering the amount of papers that are present.

 

[Racek49]

I don't have links but I've been shown conversations with people in the industry (plus circumstantial evidence, such as an identical node between Kirin 9010 and 9020a which indicates a lack of equipment change) that indicate that the SSA800i was never used in production. I'm aware that this isn't ideal and you're free to distrust what I say, but I stand by that the source I got it from is credible.

As to what's actually happening with SMEE? No idea. There's some talk in other forums that SiCarrier/Huawei poached some of the SMEE people but there's been pretty much no news.

I found a similar statement about a year ago on a link leading from Wiki .... It was in the English version and I suspect that it also referred to information from engineers from SMEE. I have no doubt that such problems must arise when an extremely complex machine is born, associated with the creation of a supply chain of thousands of completely new components and materials. The important thing is then to debug, debug, test new solutions and that will take some time. A year has passed since then and even that information cannot be old. Personally, I think that if the prototype was tested a year ago, then work on it continues and is already in the next phase.
Yes, yes, even with my education, I am now just a layman. I am just trying to find connections.

 

[tokenanalyst]

Panmeng Semiconductor has completed nearly 100 million yuan in Series A financing and is a domestic ultra-pure etching silicon material company.​

Jiangxi Panmeng Semiconductor Technology Co., Ltd. ("Panmeng Semiconductor"), a leading domestic company in ultrapure semiconductor etching silicon materials, recently completed nearly 100 million yuan in Series A financing. This round was led by Galaxy Capital, with participation from Qinling Capital, and Guangyuan Capital serving as exclusive financial advisor. The funds will be primarily used to further enhance core technology and expand production capacity, accelerating the localization of Panmeng Semiconductor's etching silicon materials and meeting the urgent global market demand for ultra-large single crystal and nitrogen-free polycrystalline materials.

Public information shows that Panmeng Semiconductor was established in 2021 and focuses on the production of semiconductor-grade silicon wafers. Its products include grinding wafers, etching wafers, and polishing wafers, which are all urgently needed for the development of the domestic semiconductor and electronics industries. The wafer sizes are 4 inches, 5 inches, 6 inches, and 8 inches. Currently, it mainly focuses on 4~6-inch silicon wafers, and will focus on 8 inches in the future.

Currently, China's domestic production rate for silicon etching materials remains low, and the high-end market has long been monopolized by foreign companies. Leveraging core breakthroughs such as independently developed large-scale temperature field control technology and nitrogen-free purification processes, Panmeng Semiconductor has become one of the few high-tech companies in China to achieve mass production of large-scale single-crystal silicon materials, breaking the monopoly of Japanese companies in the nitrogen-free polycrystalline market.

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

 

[tokenanalyst]

Xianwei Technology officially ships its first atmospheric-pressure GaN MOCVD equipment.​

On September 17th, Wuxi Firstway Technology Co., Ltd. officially shipped its independently developed EliteMO series atmospheric-pressure GaN MOCVD system to a leading compound semiconductor wafer fab in China. This marks another major breakthrough for Firstway in compound semiconductor epitaxial equipment, following the successful delivery of the BrillMO series GaN MOCVD system in June of this year. It further demonstrates the market's strong recognition of Firstway's products and technological capabilities.

The EliteMO series of metal-organic chemical vapor deposition (MOCVD) epitaxial growth equipment shipped this time boasts a wide range of applications and leading-edge process performance, suitable for the growth of GaN LDs, green to UV LEDs, GaN electronic devices, GaN-based novel structural materials, Ga₂O₃ epitaxy, and AlN epitaxy. This equipment features multi-zone resistive heating and excellent temperature uniformity, offering solutions for epitaxial growth of various materials, including multi-component and n/p-type materials, enabling the epitaxial R&D and manufacturing of advanced materials.

Please, Log in or Register to view URLs content!

 

[Racek49]

I found a similar statement about a year ago on a link leading from Wiki .... It was in the English version and I suspect that it also referred to information from engineers from SMEE. I have no doubt that such problems must arise when an extremely complex machine is born, associated with the creation of a supply chain of thousands of completely new components and materials. The important thing is then to debug, debug, test new solutions and that will take some time. A year has passed since then and even that information cannot be old. Personally, I think that if the prototype was tested a year ago, then work on it continues and is already in the next phase.
Yes, yes, even with my education, I am now just a layman. I am just trying to find connections.

Oh, sorry for the haste. Before I wrote that, the problem had been well resolved among more knowledgeable colleagues.

 

[gadgetcool5]

Huawei's deputy chairman says they have delivered "over 300" CloudMatrix 384's this year. Still, that is a very small number considering that Nvidia was confirmed to have delivered 1,500 GB200 NVL72's by April of this year already, and the GB200 NVL72 is far more powerful than the CloudMatrix 384 once you consider the better HBM, NVLink, and interconnect performance. And even that delivery is believed to be possible only because of reused dies from TSMC that have since been cut off. Morgan Stanley says only 23,000 Ascend 910Cs will be made this year at SMIC at a 5% yield. So it's ridiculously expensive and only enough for 59 CloudMatrix 384's even if all the Ascend 910Cs are used for the CloudMatrix 384.

And even if they can fix their Ascend 910C problem, they've got the HBM problem. And even if they fix the Ascend 910C and HBM problem, they've got the toughest problem of all which is that Mindspore's ecosystem is underdeveloped and immature compared to CUDA. So they can announce all the ambitious plans they want but the question is, what have they achieved so far? The reality is that China is being starved of AI training power.