Chinese semiconductor industry

The first quantum chip production line in China


Equipments I'm able to identify from the video

Mask Aligner from SUSS Germany


Thermal process machine from Ultech Korea


ALD machine from Yunmao China


Heating Microscope from Hesse-instruments Germany


Microscope from Olympus Japan

theorlonator said:

He said a lot of things in that article. I'm curious how he can claim what he claims. For instance he said the yield for SMIC's deAmericanized line is only 25%. Really?

Click to expand...

I respond with humor because it's almost a joke article. How would he know any of that, as yield is proprietary data? either he's a liar or will have a lawsuit incoming soon. What is the process node and die size for 25% yield? Smaller process and bigger die usually has lower yield. So any yield number comes with conditions. Samsung's yield issues for instance were highly specified down to which exact chip was being produced.

He then claims that AMEC, NAURA, etc are all lying about their 5-14 nm etchers as he says they can only etch 5 nm in simple conditions. How would he know that unless he was hands on in evaluating them, and even then it contradicts their real life usage. And he admits he has been out of the industry for a few years, so how would he know the latest news? NAURA grew exponentially in the past few years.

He finally goes on this unhinged rant about "wolf warriors", how we need to be "humble", how we are nothing without "the world", how he reminisces about the days where Taiwanese and Japanese would manage him, etc. He also has a huge rant in there about how China's science and engineering academia has no idea what they're doing. Does this sound like some objective analysis to you???

proelite said:

The US will add Xiaomi, Tencent, Lenovo, etc to the entity list eventually. There is no avoiding fate for the traditional big tech in China. They'll all get Huaweied.

What's required now is leapfrogging and pivoting across the barrier that is the US export controls. Subsidizing the domestic semiconductor industry beyond just able to provide mature tech is a waste of money.

Click to expand...

Dirty little secret is that companies like bytedance, tencent, alibaba, lenovo etc are heavely invested by western companies/banks/funds, the likely hood of them being ban is less in ordinary circumstance than pure Chinese companies like huawei or companies with lower western investments.

Part of the ban besides knocking Chinas tech and economic ambition , it to also prevent a Chinese attempt to invade Taiwan. As long as there is no war, China will get the consumer chips. China just needs to hide and bide for more time at this point to be self sufficent. IMO, there may have been some miscalulation with "coming out" too soon especially with Huawei then seeing their entire supply chain wrecked.

gadgetcool5 said:

Loongson's newer processors use 28nm and under nodes and Moore Threads use 12nm.

Those processors are made using American, Japanese and Dutch equipment. Therefore they cannot supply Huawei and are vulnerable to sanction.

Huawei cannot work with Longsoon or Moore Threads. Huawei cannot work with any well known domestic architecture. Huawei also cannot work with SMIC because that company's production lines use Dutch and Japanese equipment. What you need is the following: All Chinese equipment including lithography machines (Currently such production lines do not exist for mass production, as far as I know). All Chinese CPU architecture and IP (Must be manufactured on 90nm and lower). All Chinese software ecosystem to support it (Still in infancy, not sure if Chinese OS can take off). And you need it to be competitive in terms of performance. This day has been known and coming since 2020. The mountain is extremely high to climb. It can be climbed, but it will take a long time and the question is whether the will exists. It might be easier for China to pursue leapfrog technologies instead.

Click to expand...

Loongson and SMIC are on US blacklist along with Huawei already if I'm not mistaken

daifo said:

Dirty little secret is that companies like bytedance, tencent, alibaba, lenovo etc are heavely invested by western companies/banks/funds, the likely hood of them being ban is less in ordinary circumstance than pure Chinese companies like huawei or companies with lower western investments.

Part of the ban besides knocking Chinas tech and economic ambition , it to also prevent a Chinese attempt to invade Taiwan. As long as there is no war, China will get the consumer chips. China just needs to hide and bide for more time at this point to be self sufficent. IMO, there may have been some miscalulation with "coming out" too soon especially with Huawei then seeing their entire supply chain wrecked.

Click to expand...

There was no coming out. You can't hide when they're looking for you. Huawei was on their shit list as of 2009, how the fuck is that Huawei's problem for "coming out"? The only way Huawei doesn't "come out" is for Huawei to be an unimportant company with no useful IP or market share. Extend that to China as a whole. So how the hell is that better than being sanctioned?

You didn't assign responsibility for escalation correctly thus your conclusion is wrong.

The sequence of events is not "Huawei boasted, US sanctioned, thus Huawei is at fault for boasting. The solution is to submit." It is "Huawei did nothing except normal commercial activities. US got paranoid and struck first with sanctions. Thus it is US fault and the solution is to not only recover but to strike back."

FairAndUnbiased said:

There was no coming out. You can't hide when they're looking for you. Huawei was on their shit list as of 2009, how the fuck is that Huawei's problem for "coming out"? The only way Huawei doesn't "come out" is for Huawei to be an unimportant company with no useful IP or market share. Extend that to China as a whole. So how the hell is that better than being sanctioned?

You didn't assign responsibility for escalation correctly thus your conclusion is wrong.

The sequence of events is not "Huawei boasted, US sanctioned, thus Huawei is at fault for boasting. The solution is to submit." It is "Huawei did nothing except normal commercial activities. US got paranoid and struck first with sanctions. Thus it is US fault and the solution is to not only recover but to strike back."

Click to expand...

If boasting can get someone sanctioned then a certain supapowa would be the most heavily sanctioned country on the planet.

theorlonator said:

Did anyone see this post? He's extremely pessimistic:

Please, Log in or Register to view URLs content!

Click to expand...

IMHO. tools, tools, tools nothing but tools. Between Chinese liberals and gradualist like this author is why Chinese companies are in this situation. semiconductor manufacturing equipment, EDA and semiconductor materials do not develop in vacuum, they need a virtuous cycle to grow and develop, is necessary for local companies to adopt local tools. The government could had forced the use of SMEE frontend machines to spur their development, at least their I-line and Krf line, at least in certain nodes, regardless the opinion of ASML, but they decided for fabs to select their tools on their own or use the bidding system to do that.
I agree that a lot of money went to sectors of the industry like chip design and others companies that then used that money to enrich foreign and especially US companies instead of fostering a domestic semiconductor ecosystem.
I disagree on that academic part, yes in the past was difficult for Chinese academic institutions collaborate with the industry and monetize their inventions. But on the years I have been following this I have never seen so much collaboration between research institutions and the industry as since 2018, they getting more patents, collaborating with establish domestic companies and they are setting more companies on their own to support the industry.
For SMIC maybe is 25% but the localization rate of the overall China semiconductor industry to this day is closer to 35% but whatever, in 2020 china localization rate was just 7%, for SMIC probably even lower.
So the glass is half full or half empty?

tonyget said:

Mask Aligner from SUSS Germany

Click to expand...

The SSB300 would be too much of an overkill for this product. Looks like they are doing a chip at a time.


maybe for this kind of application would be better to go maskless to have the flexibility of no using a contact mask.

theorlonator said:

Did anyone see this post? He's extremely pessimistic:

Please, Log in or Register to view URLs content!

Click to expand...

Here some interesting bits (translated with google, sorry)

以目前推进国产设备最积极的中芯国际为例，2021年的新厂中芯京城国产化率只有25%，这还是在近几年花大力气帮所有Vendor做认证的情况下的现状——因为要采购国产设备，得先经过大量测试，认证合格才能采购。另一方面，这两年几乎所有Fab都将认证标准大幅度降低，2023年的新招标，在政府的要求之下，最多也只能将国产化率拉到45~50%，并且这50%都属于相对容易攻克的部分。
Take SMIC, which is currently the most active in promoting domestically produced equipment, as an example. The localization rate of the new factory in Beijing in 2021 is only 25%. Because if you want to purchase domestic equipment, you have to go through a lot of tests and pass the certification before purchasing. On the other hand, in the past two years, almost all fabs have significantly lowered the certification standards. Under the government's request, the new bidding in 2023 can only raise the localization rate to 45-50% at most, and this 50% They are all relatively easy to overcome.

(as it was obvious, Chinese government went in...IMHO it should had done it well before than just the last 2 years)


良率拉升需要时间，以中芯国际为例，其28nm含美线良率在85%左右，实现25%国产化率的非美线良率不到七成，最好的结果是可以拉到75% It takes time to increase the yield rate. Taking SMIC as an example, its 28nm yield rate with American lines is about 85%, and the yield rate of non-American lines that achieves 25% localization rate is less than 70%. The best result is that it can be pulled to 75%

(As I read it, without US suppliers the 28nm line, can reach 75% yield instead of 85%. This is not such a bad result at all, it means that the de-americanized line works, regarding his worries on the production costs I don't see it as a sensible point in this moment and for the next several years too).



中芯国际于去年底采购的30多台用于28nm以上的NXT1970以及1965CI型号DUV浸润式光刻机，将于2024年陆续到货，这30多台光刻机也就是中芯京城，中芯东方，深圳与天津四个新建Fab所需要的设备。More than 30 NXT1970 and 1965CI model DUV immersion lithography machines purchased by SMIC at the end of last year will arrive in succession in 2024. These more than 30 lithography machines are SMIC Beijing, SMIC The equipment needed for the four new Fabs in Dongfang, Shenzhen and Tianjin.




作者记得2001年第一次踏上祖国土地
The (article) author remembers setting foot on the motherland for the first time in 2001

(maybe he is not from mainland China)

Shanghai Microsystems Institute has made new progress in the development of phase-change memory with extreme operating speed​

Today's computer systems employ a hierarchical storage architecture: cache, memory, and flash memory. The closer to the CPU , the higher the demand for memory storage speed. For example, the speed of memory is at the nanosecond level, while the cache needs to be at the picosecond level. As a strong competitor of next-generation memory, the speed of phase-change memory determines its application field, and the speed of phase-change memory is mainly determined by the crystallization speed (writing speed) of phase-change materials. Studies have shown that the worse the thermal stability of phase change memory, the faster the crystallization speed, and elemental antimony ( Sb ), which is the phase change material with the worst thermal stability known so far, may have the fastest operating speed.
Through molecular dynamics calculations, the research team of Song Zhitang and Zhu Min of Shanghai 　　Institute elemental antimony can nucleate from the amorphous structure and further crystallize completely within 120 ps . By preparing 200 nm , 120 nm and 60 nm T -type low-level device elemental antimony phase-change memory devices, it is found that as the device size decreases, the speed of the elemental antimony phase-change memory is faster. The fastest writing speed of the 200 nm elemental antimony device is 359 ps (see Figure 1 ), and when the device size is shrunk to 60 nm , the writing speed is ~242 ps, which is nearly 100 times faster than that of the traditional Ge 2 Sb 2 Te 5 ( 20 ns ). By comparing the speed of the reported phase change memory (see Figure 2 ), the speed of the single substance Sb device is significantly faster than the traditional Sb-Te , Ge-Te and Ge-Sb-TeBased phase-change memory, its ~242 ps operation speed is the limit of current phase-change memory speed. This result shows that phase change memory is expected to have the potential to replace memory and even cache by selecting appropriate phase change materials.