I guess you mean mm^2.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
ZTE plans to complete the spin-off of ZTE Microelectronics by the end of September
According to micronet news, the source said that ZTE is seeking to spin off ZTE Microelectronics, which was originally planned to be completed by the end of September. It is reported that ZTE Microelectronics will consider financing in the form of an independent company, and may seek to go public in the future.
As a holding subsidiary of ZTE, ZTE Microelectronics was established in 2003, formerly the IC design department of ZTE. ZTE Microelectronics focuses on the research and development of ICT chips, providing core chips and solutions in the fields of wireless communication, broadband access, optical transmission, routing and switching, etc. In addition to providing support for ZTE's own chip research, it also began to sell chip products and design solutions. Program.
Sir most of them are foreigners or locals?
locals and foreigners
Yes, it is.
They specialize on sparse neural networks. It is like a normal net, but with the majority of weights set to zero. This allows a huge speed up compared to the standard dense networks, using tailored special algorithms for sparse matrix computation.
If you look at the ML bench table, you will see that they beat NVIDIA on the special case of sparse network, so maybe their hardware is not so powerful like NVIDIA in the general case, but they won out of a smarter approach.
It's telling that ASML China is way bigger than SMEE in China!
It's clear they strongly want to remain in China for the future, and I think they know very well that if they give up on DUV banning their future in China is doomed.
On the other hand it keeps getting clearer every day that the ban of ASML litho machine is the only real weapon by US to strongly impact Chinese growth for some years to come. At the moment also banning of TSMC on advanced nodes is a very powerful one, but as soon as the new SMIC sub 14nm fab will come online, also this one will become moot.
I can only imagine the kind of pressure Holland government and ASML have to deal with! If ASML would have been American, it would have been already game over!
Point well-taken. It is hard to keep up with the flurry of news coming out of China's tech scene. Special thanks to you for writing such detailed comments and putting the news in context, which isn't always the case when you read the media articles.
there is still a lot of room for domestic machines.... I'm not sure how local companies will guess this big wave... but for now let ASML be comfortable with it... from November this year domestic lithography don't have to worry about sales since 2023
yes, sorry. m^2 would be pretty large. @latenlazy pointed to that the drop off in # of wafers from N+1 to N+2 might not as drastic and the sales price increase might be higher. I will just to 400k per year for N+1 and 300k per year for N+2
So, doing another calculation, for desktop chips if Phytium is selling D2000 right now for $70 per chip and SMIC is getting $20 per chip using their 14 nm process, then their yearly revenue would be the following assuming a 80% yield on the process
$20*240*600k=2.88 billion
if they can sell a new N+1 CPU for $120 per chip and SMIC is getting $40 per chip using N+1 process, then the revenue would be following assuming a 70% yield on the process.
$40*210*400k=3.36 billion
And if they can sell S5000 server CPU for $400 per chip and SMIC is getting $120 per chip using N+1 process, then revenue would be the following if we assume like a 60% yield on the process. I'm assuming yield lower on larger chips.
$120*70*400k=3.36 billion
And if Alibaba can sell Yitian-710 for $800 per chip and SMIC is getting $200 per chip using N+2 process, then revenue would be the following if we assume like a 50% yield on the process
$200*60*300k=3.6 billion
For smartphones, let's say if Huawei/HiSilicon sells its N+2 CPU for $120 and SMIC is getting $40 per chip and the yield is 2/3 (given smaller size than desktop chip)
$40*400*300k= 4.8 billion
It would seem to me that smartphones is far and away the most lucrative market if they can achieve high enough yield. 300k*400 -> 120 million a year. Seems like if they can fully ramp up all 3 fabs and they can all do 300k wafers a year on N+2, then they can satisfy the low to medium domestic smartphone market + all the other requirements. Smartphone would require 2 fabs and all the other chips can be produced in the other fab. If I'm SMIC, I would try to get into the smartphone market as soon as I can, because that's where the money is. In fact, I don't know why they would build so many new "28 nm fabs" with high production capacity if they weren't trying to get into the more competitive smartphone market. If I'm SMIC, I would let other chip makers take over the lower end market.
- Status