Chinese semiconductor industry

so this actually probably will help Huawei, but definitely chinese phone companies as a whole. SOEs & other national security industries. the question is whether they can actually enforce this. It's been requested before and never actually carried out

hvpc said:

It's possible but not realistic.

Too many variables impacting the final economic penalty from using DUV to build 5nm. No one had succeeded in using DUV to build 5nm chips. All the claim on social media that DUV can be used to make 5nm are all theoretical (on paper).

We can calculate the CapEx delta between using DUV vs. EUV for 5nm. We'll find cost behind using EUV would be cheaper than DUV. But the biggest economic delta between DUV & EUV will come from wafer yield or good-dies-per-wafer performance between the two. Since no one had succeeded thus far in commercializing 5nm using only DUV far, there's really no point in making the comparison.

These questions are akin to asking if it is possible to swim (instead of via airplane) across the Pacific from the US to China & how much more expensive it would it be.

Click to expand...

what about the SMIC N+2 used in kirin 9000s? can we estimate the cost increase compared with players using EUVs, like TSMC?

Jingsheng Announced on the evening of August 28 , Nanjing Jingsheng Equipment Co., Ltd. recently signed the "[Silicon Carbide Furnace] Equipment Sales Contract" with customer J. The estimated total contract price (including tax) is 120,350,650.00 yuan, which will be implemented in three batches, of which the quantity, model, price and delivery time of the second and third batches of equipment shall be subject to the order signed by both parties.



The company is a supplier of special equipment for semiconductors, mainly engaged in the research and development, production and sales of crystal growth equipment, and provides customized semiconductor-grade monocrystalline silicon furnaces, silicon carbide single crystal furnaces and other crystal growth equipment to semiconductor material manufacturers and other material customers. product.

According to the development trend of downstream technology and applications, the company has been the first to invest in the research and development of 4-6 inch conductive silicon carbide single crystal furnaces since 2018, and achieved the first product sales in 2019, and the products will be put into downstream silicon carbide power devices in batches in 2020 ( Conductive substrate) application field verification and application. Based on the technical capabilities of the conductive silicon carbide single crystal furnace, the company will complete the research and development, improvement and finalization of the 6-inch semi-insulating silicon carbide single crystal furnace in 2020. In 2020, the company's semi-insulating silicon carbide single crystal furnace will achieve the first supply and product verification to Tianyue Advanced, a leading domestic semi-insulating substrate manufacturer.

With the continuous development of the domestic silicon carbide industry, the company has successively realized the mass supply of products to downstream manufacturers such as Sanan Optoelectronics, Tony Electronics, and Zhejiang Jingyue, and continued to promote batch products with more than ten downstream silicon carbide manufacturers such as BYD Sales business, established a good and stable business cooperation relationship with the above customers.

Psyclonus said:

what about the SMIC N+2 used in kirin 9000s? can we estimate the cost increase compared with players using EUVs, like TSMC?

Click to expand...

As I understand it TSMC 7nm is comparable to Intel’s most advanced process. Intel is also using DUV multi patterning, not using EUV due to cost. In other words, in principle, it could be competitive.

Funny how MSM downplays the Huawei 7nm breakthrough when technically speaking Intel is on the same or worse level.

Psyclonus said:

what about the SMIC N+2 used in kirin 9000s? can we estimate the cost increase compared with players using EUVs, like TSMC?

Click to expand...

imo, it's not worth the effort to estimate these things. Based on my calculations, both SMIC & Huawei will do well here and you shouldn't worry about it.

tokenanalyst said:

Jingsheng Announced on the evening of August 28 , Nanjing Jingsheng Equipment Co., Ltd. recently signed the "[Silicon Carbide Furnace] Equipment Sales Contract" with customer J. The estimated total contract price (including tax) is 120,350,650.00 yuan, which will be implemented in three batches, of which the quantity, model, price and delivery time of the second and third batches of equipment shall be subject to the order signed by both parties.

The company is a supplier of special equipment for semiconductors, mainly engaged in the research and development, production and sales of crystal growth equipment, and provides customized semiconductor-grade monocrystalline silicon furnaces, silicon carbide single crystal furnaces and other crystal growth equipment to semiconductor material manufacturers and other material customers. product.

According to the development trend of downstream technology and applications, the company has been the first to invest in the research and development of 4-6 inch conductive silicon carbide single crystal furnaces since 2018, and achieved the first product sales in 2019, and the products will be put into downstream silicon carbide power devices in batches in 2020 ( Conductive substrate) application field verification and application. Based on the technical capabilities of the conductive silicon carbide single crystal furnace, the company will complete the research and development, improvement and finalization of the 6-inch semi-insulating silicon carbide single crystal furnace in 2020. In 2020, the company's semi-insulating silicon carbide single crystal furnace will achieve the first supply and product verification to Tianyue Advanced, a leading domestic semi-insulating substrate manufacturer.

With the continuous development of the domestic silicon carbide industry, the company has successively realized the mass supply of products to downstream manufacturers such as Sanan Optoelectronics, Tony Electronics, and Zhejiang Jingyue, and continued to promote batch products with more than ten downstream silicon carbide manufacturers such as BYD Sales business, established a good and stable business cooperation relationship with the above customers.

Please, Log in or Register to view URLs content!

Click to expand...

yep, Crystal growth sold quite a few to Sanan & BYD. I don't think they are the only one though. i think NAURA also makes them

tokenanalyst said:

The biggest irony is if SMIC had never been sanctioned, I think that Huawei would not have been able to accomplish this feat. Before export controls Chinese companies were notorious for throwing other Chinese companies under the bus.

Click to expand...

I think we actually need to be very careful about what Huawei did here that was special. Not enough people are realizing it & pointing to wrong places

But beyond that, SMIC getting yield of Kirin 9000S to a good place is the most immediate goal. Big difference between the output of a 40% yield process vs a 75% yield process that I think they need to get to.

Look, if they can get even 40 million phones next year using their own chip, the yield will already be really good

hvpc said:

this would be true if SMIC have enough capacity to support all the domestic advanced IC needs & displace the need for existing+additional wafer capacity at tsmc.

folks, celebrate the Kirin9000s breakthrough, but we got to stop counting the chicken before they hatch. China still has long way to go before catching up to the western controlled ecosystem supply chain.

Click to expand...

I think it's fine for people to celebrate because it's been a tough year for advanced chipmakers in China, but we just need to stay on topic, which unfortunately some of these posts are just too speculative & celebatory

siegecrossbow said:

Funny how MSM downplays the Huawei 7nm breakthrough when technically speaking Intel is on the same or worse level.

Click to expand...

Intel 4 is already in production and they are on their way to Intel 3 production soon.

hvpc said:

It's possible but not realistic.

Too many variables impacting the final economic penalty from using DUV to build 5nm. No one had succeeded in using DUV to build 5nm chips. All the claim on social media that DUV can be used to make 5nm are all theoretical (on paper).

We can calculate the CapEx delta between using DUV vs. EUV for 5nm. We'll find cost behind using EUV would be cheaper than DUV. But the biggest economic delta between DUV & EUV will come from wafer yield or good-dies-per-wafer performance between the two. Since no one had succeeded thus far in commercializing 5nm using only DUV far, there's really no point in making the comparison.

These questions are akin to asking if it is possible to swim (instead of via airplane) across the Pacific from the US to China & how much more expensive it would it be.

Click to expand...

If Huawei use Qualcomm chips that is made with high yield, how does it benefit Huawei? Qualcomm will not suddenly sell it at a low profit margin because the yield is good.

Both TSMC and Qualcomm have a high profit margin and they are driven by high end products.

It is true that EUV is needed for N5 for TSMC because the defect t density is lower than N7 which was 0.09 5q after high volume manufacturing.

Suppose using DUV to build 5nm is 3x more complex than, so the defect density is 0.3. so the yield is lower, so what? Chinese companies never live on the profit margin that the US firms are.

It wouldn't make sense for TSMC because it has EUV, it would make sense on China for now because it doesn't. It will not be as profitable, but it will deny revenue to US and TSMC.

so this is the interesting to watch
Huawei will have product announcements over this next week from new smart screens to ear buds to watches to IoT (they got a new smart air purifier) which all needs at least 28nm (I would imagine in the case of air purifier), but in many cases 14nm. So, it's not like SMIC can just switch its 14nm production all to 7nm even if it wants to

At the same time, it needs to prepare for the very likelihood that there is a full ban on chip sales. Now of course, right now, it uses SD 888 for its pad & Intel CPUs for notebook & cloud, but those will run out by 2025 probably. I'm assuming for example the 40m SD 8 that it ordered this year will be used on 4G phone sales & pads (I assume some of them are for pads) over next year or 2. But most of the phone SoCs needed are 5G. So for next year, they need to have probably 40m Kirin 9000 if they want to sell 60m phones

so for 2025, SMIC needs to be able to produce probably 60m phone SoC, 5-10m desktop CPU, 2m server CPUs & half million GPUs. And yield for GPUs/server CPUs and desktop CPUs will be a lot lower than phone SoC

So, we really the domestic tools to get validated & SMIC production to be ramped up.

Also, keep in mind, Huawei actually cannot use up 100% of SMIC's Finfet production despite what some people think