Opinionated - China Chipping Away to Semiconductor Dominance

[j20blackdragon]

America’s tough talk has backfired, prompting Huawei to redouble efforts to wean itself off American technology. Although such claims are hard to verify, Tim Danks, a Huawei executive, has said that of the 600,000 base stations the firm has shipped to mostly Chinese carriers, 50,000 had no American parts. Engineers who recently took apart Huawei’s top-end smartphone identified only a few American-designed chips.

Left to their own devices, American carriers have put up 10,000 5g base stations. Covid-19 lockdowns will slow installations in the West, even as China eases its own curbs now that its epidemic is apparently under control. Chinese carriers, which boast 150,000 base stations, want more than 1m across 330 cities by the end of 2020.

https://www.economist.com/business/...not-want-china-to-dominate-5g-mobile-networks

 

[Paul2]

40+ nm is too backwards IMO. it excludes important mainstream applications that aren't processors such as FPGA, memory and RF. at least for memory, there is no substitute for <14 nm tools. maybe for FPGA and RF you can just make shittier products I guess.

80%+ of Chinese IC volume is 90nm+. There are tons, and tons of ICs other than chips you buy for $100+, in fact the <40nm stuff only takes single digit percentage of wafer shipments for Chinese electronics.

The choice is a no brainer. <10% of wafer volume for grabs, or >90%? There is still soo much unmet demand left.

So far, the state was only ready to dish out grants for vainglorious pursuit of bleeding edge, without any meaningful end goal in mind. So much money went into the drain... ah

IMHO, the electronics section at MOFCOM don't see the electronics industry beyond a dozen or so showpiece companies. Achieving industrial independence requires the whole of industry to work, and not only the 1% cream on top that keeps going broke upon every market shift.

Unigroup can keep spraying state money around on fancy projects for which they collect laurels, but the ultimate determinator of success here would be people buying their stuff, and I am not seeing that.

the other thing is that EUV's problems only show up at leading edge process nodes. You can still use EUV processes for non-leading edge like 14+ nm.

You can, but why? The best what TSMC gets out of NXE3400 is 200-300 wafers per hour. This is what I heard. An immersion DUV tool will easily have output it thousands. Even if you save on multiple patterning passes, your throughput will be few times smaller. TSMC only uses EUV for some device layers and M0, everything else is done on immersion tool.

finally, EUV is still new, while refining older processes are going to be more difficult because of so many incumbents.

You just said what is the crux of the matter here. Chinese companies have to go abroad to fab chips on mature, mainstream processes that makes the most of fab service supply. In other words, somebody well moneyed like Huawei or Unisoc can do an expensive tapeout at SMIC or Huali on a moderately advanced process at above market price, but the rest of us have to go to Taiwan even for the most basic projects.

Through internal restructuring, the subsidiary, originally named BYD Microelectronics Co Ltd, now covers the development, production and sale of power semiconductors, intelligent control products, intelligent sensors and optoelectronic semiconductors. It boasts an integrated industrial chain with chip design, wafer manufacturing, assembly, testing and downstream applications.

The company said that with the restructuring, it will increase its supply of insulated-gate bipolar transistor (or IGBT) semiconductor technology to the industry. The IGBT semiconductor technology is regarded as a core technology of new energy vehicles (NEV).

Song Jin, a senior automobile analyst at Beijing-based market consultancy Analysys, said: "Semiconductors are important hardware for NEV production. BYD's restructuring of its subsidiary offers the subsidiary more autonomy, and promotes the company to better operate its semiconductor business."

This is the real deal. I knew a few BYD engineers going to industry events in Shenzhen who talked specifically of their problems with power electronics. Was shown few photos of a their briefcase sized inverter pack, and said it was their pain.

Our high speed trains use CRRC's own IGBTs (foreign high speed train designs all use plain iron transformers,) but there is not much to them other than them being fabbed in China.

 

[FairAndUnbiased]

80%+ of Chinese IC volume is 90nm+. There are tons, and tons of ICs other than chips you buy for $100+, in fact the <40nm stuff only takes single digit percentage of wafer shipments for Chinese electronics.

The choice is a no brainer. <10% of wafer volume for grabs, or >90%? There is still soo much unmet demand left.

So far, the state was only ready to dish out grants for vainglorious pursuit of bleeding edge, without any meaningful end goal in mind. So much money went into the drain... ah

IMHO, the electronics section at MOFCOM don't see the electronics industry beyond a dozen or so showpiece companies. Achieving industrial independence requires the whole of industry to work, and not only the 1% cream on top that keeps going broke upon every market shift.

Unigroup can keep spraying state money around on fancy projects for which they collect laurels, but the ultimate determinator of success here would be people buying their stuff, and I am not seeing that.
You can, but why? The best what TSMC gets out of NXE3400 is 200-300 wafers per hour. This is what I heard. An immersion DUV tool will easily have output it thousands. Even if you save on multiple patterning passes, your throughput will be few times smaller. TSMC only uses EUV for some device layers and M0, everything else is done on immersion tool.You just said what is the crux of the matter here. Chinese companies have to go abroad to fab chips on mature, mainstream processes that makes the most of fab service supply. In other words, somebody well moneyed like Huawei or Unisoc can do an expensive tapeout at SMIC or Huali on a moderately advanced process at above market price, but the rest of us have to go to Taiwan even for the most basic projects.


This is the real deal. I knew a few BYD engineers going to industry events in Shenzhen who talked specifically of their problems with power electronics. Was shown few photos of a their briefcase sized inverter pack, and said it was their pain.

Our high speed trains use CRRC's own IGBTs (foreign high speed train designs all use plain iron transformers,) but there is not much to them other than them being fabbed in China.

Trying to type from phone at work, so thoughts are disjointed. My viewpoint is very different than yours. I work at a process machinery and chemicals OEM. I honestly don't even think about >90 nm stuff, that stuff seems to be a solved problem already. At conferences I go to nobody talks about those tools, because even used tools after refurbishment will work.

Right now our customers including ASML, Lam, etc. are mostly thinking of leading edge, and spec out their equipment parts and chemicals accordingly, even if they don't solely sell products for the leading edge. our R&D efforts are such that our new products are fit for <14 nm and EUV, even if they can be used elsewhere.

IMO MOFCOM has no duty to support purely commercial ventures. if you only make what an immature market wants right now, how can you get past the middle income trap and grab the commanding heights of technology? If private industry wants to backfill trailing edge tech, let them do it with their own money. Government should focus money on leading edge tech and R&D.

If there is truly so much money in trailing edge ICs then the private sector can take care of it.

 

[j20blackdragon]

80%+ of Chinese IC volume is 90nm+.

I honestly don't even think about >90 nm stuff, that stuff seems to be a solved problem already. At conferences I go to nobody talks about those tools, because even used tools after refurbishment will work.

Nobody cares about >90 nm because there's no money there.

 

[Paul2]

seems to be a solved problem already

The 80% of wafer volume figure says it's not. That's obvious.

Nobody cares about >90 nm because there's no money there.

You are a business with less than 10 millions in bank? Then, you will care. Count Chinese companies who can afford a tapeout on 14nm (cost 800 millions and more.) There are only two now, with 3rd unlikely coming under current circumstances...

Right now our customers including ASML, Lam, etc. are mostly thinking of leading edge, and spec out their equipment parts and chemicals accordingly, even if they don't solely sell products for the leading edge. our R&D efforts are such that our new products are fit for <14 nm and EUV, even if they can be used elsewhere.

Yes, if you are tool maker, the cash is where the clients with a lot of cash are, but if you are an IC maker, it's a lost game.

The only two ways an IC company with <CNY 100 in capital can realistically make money in China is by going after stable yield cash crop products like analog, and commodity parts, or watching the market and trying to spin new niche products into mainstream.

Have anybody of you here ever shipped an IC? I did.

If there is truly so much money in trailing edge ICs then the private sector can take care of it.

This is happening, but really, really slow. CanSemi took close to 5 years to spin up. Few other domestic private fab projects are still hanging by a thread. HonHai had 6 years long epopee with buying land for the plant.

CanSemi is making bucket loads of money now, but without much competition they didn't move the price in the market in China by much.

if you only make what an immature market wants right now, how can you get past the middle income trap and grab the commanding heights of technology?

I see it like this: there are 2 microelectronics industries existing in parallel in China now. One is the real one, stuck making bottom feeder products, and another one exists in the bubble of state ran projects, where people are fine paying n times the market rate for expensive tapeouts.

Yes, the bosses of state companies can spend many billions to build that EUV <7nm fab for a checkbox on the resume, but what it would really change if there would be nobody in China to use it? Besides Unisoc, few similar state clients, and, possibly, Huawei.

What you read in economics textbooks is not how the real world works. Middle income trap, or not, you cannot progress further until you cover the basics well. I think this is very much the reason why middle income traps happen in the first place, the choice of pie in the sky project over what will actually make money now.

 

[FairAndUnbiased]

The 80% of wafer volume figure says it's not. That's obvious.

You are a business with less than 10 millions in bank? Then, you will care. Count Chinese companies who can afford a tapeout on 14nm (cost 800 millions and more.) There are only two now, with 3rd unlikely coming under current circumstances...

Yes, if you are tool maker, the cash is where the clients with a lot of cash are, but if you are an IC maker, it's a lost game.

I see it like this: there are 2 microelectronics industries existing in parallel in China now. One is the real one, stuck making bottom feeder products, and another one exists in the bubble of state ran projects, where people are fine paying n times the market rate for expensive tapeouts.

Yes, the bosses of state companies can spend many billions to build that EUV <7nm fab for a checkbox on the resume, but what it would really change if there would be nobody in China to use it? Besides Unisoc, few similar state clients, and, possibly, Huawei.

What you read in economics textbooks is not how the real world works. Middle income trap, or not, you cannot progress further until you cover the basics well. I think this is very much the reason why middle income traps happen in the first place, the choice of pie in the sky project over what will actually make money now.

Is demand really a problem? SMIC doesn't only offer foundry service for Chinese market. Their total utilization is 94%. Clearly they don't have a problem with maintaining demand. They also produce for heavyweights like Qualcomm and TI. And even with poor demand, you build it, the demand will come, as long as you offer something technically new. Even if it's from the government.

Look at Naura. They are shipping relatively advanced tools and growing their profits 30%. There's no reason to hold yourself back. Build it, and customers come. That's been my experience with new tools. It'll take a few years to ramp up, but so what? If government doesn't support that, why even care about semiconductor? Just make shoes, ice cream, etc. Just accept slavery.

Middle income traps happen because of low innovation, lack of R&D and too much focus on short term profits. Look at all the countries falling into the middle income trap like Brazil, Thailand, Indonesia, etc. What do they have in common? Low innovation, lack of R&D, no guidance. They get comfortable making shit tier products and selling the natural resources of their people, resources needed to secure our existence and prosperity.

Let private enterprise sort themselves out. In running industrial policy, you can't cater to them. That is not the responsibility of the government. The government should invest in capability if private enterprise is unable to. That is literally the point of state sponsored research. There are some things you can't buy regardless of the price. What good is a few more pennies then?

 

[Han Patriot]

Is demand really a problem? SMIC doesn't only offer foundry service for Chinese market. Their total utilization is 94%. Clearly they don't have a problem with maintaining demand. They also produce for heavyweights like Qualcomm and TI. And even with poor demand, you build it, the demand will come, as long as you offer something technically new. Even if it's from the government.

Look at Naura. They are shipping relatively advanced tools and growing their profits 30%. There's no reason to hold yourself back. Build it, and customers come. That's been my experience with new tools. It'll take a few years to ramp up, but so what? If government doesn't support that, why even care about semiconductor? Just make shoes, ice cream, etc. Just accept slavery.

Middle income traps happen because of low innovation, lack of R&D and too much focus on short term profits. Look at all the countries falling into the middle income trap like Brazil, Thailand, Indonesia, etc. What do they have in common? Low innovation, lack of R&D, no guidance. They get comfortable making shit tier products and selling the natural resources of their people, resources needed to secure our existence and prosperity.

Let private enterprise sort themselves out. In running industrial policy, you can't cater to them. That is not the responsibility of the government. The government should invest in capability if private enterprise is unable to. That is literally the point of state sponsored research. There are some things you can't buy regardless of the price. What good is a few more pennies then?

Exactly, almost all mid income trapped countries are highly dependent on foreign tech, China can be considered middle income now, but we do have relative advanced technology. I believe we can cross this threshold and reach high income around 2050.

 

[S10]

Exactly, almost all mid income trapped countries are highly dependent on foreign tech, China can be considered middle income now, but we do have relative advanced technology. I believe we can cross this threshold and reach high income around 2050.

One step at a time:

2025中国制造
2035中国智造
2045中国创造

 

[ILC]

Yes, the bosses of state companies can spend many billions to build that EUV <7nm fab for a checkbox on the resume, but what it would really change if there would be nobody in China to use it? Besides Unisoc, few similar state clients, and, possibly, Huawei.

What you read in economics textbooks is not how the real world works. Middle income trap, or not, you cannot progress further until you cover the basics well. I think this is very much the reason why middle income traps happen in the first place, the choice of pie in the sky project over what will actually make money now.

I think that will change fast.

This is out. US new rule will try block Semiconductor equipments to China even for civilian usage. I saw a number of US articles mentions Huawei switching 14nm to SMIC. So this new rule will hit everyone in China. Not just Huawei.

I think this is good. Chinese firms usually don't do it unless being forced upon and got no other choices or they will always take the path of least resistance.

https://www.sinodefenceforum.com/t/...ithography-equipment.8493/page-69#post-600727

 

[JSCh]

DRAM appeared on CXMT website and officially on sales :-

View attachment 609643​

China-based ChangXin Memory's 19nm DDR4 chip capacity set to doubled by end of Q - cnTechPost
2020-04-30 20:20:43 GMT+8 | cnTechPost

China-based ChangXin Memory's 19nm DDR4 chip capacity set to doubled by end of Q2

ChangXin Memory Technologies' 19nm wafer chip capacity will increase from 20,000 wafers/month to 40,000 wafers/month by the end of the quarter, Mydrivers.com said, citing industry updates.

Not only that, but ChangXin's 17nm process chip is on track to be mass-produced by the end of the year, the report said.

ChangXin officially mass-produced its 10nm (10nm~19nm) DDR4 memory chips (8Gb) in September last year.

ChangXin currently has a 12-inch wafer fab with a total capacity of 120,000 wafers/month.

This week, ChangXin reached an agreement with Rambus to license the latter's DRAM technology.

Earlier, ChangXin acquired more than 10 million technical documents and 2.8TB of data related to DRAM from Qimonda through an agreement with Polaris Innovations, a wholly owned subsidiary of WiLAN.

ChangXin was launched in May 2016 in Hefei, Anhui Province, with a total investment of 150 billion yuan, specializing in the development, production and sales of DRAM memory.

 

[Paul2]

Look at Naura.

Are you with them?

 

[FairAndUnbiased]

Are you with them?

No, I work for another multinational on process equipment/chemical R&D. we supply process chemistry, coating services and ultraclean components.

 

[Paul2]

No, I work for another multinational on process equipment/chemical R&D. we supply process chemistry, coating services and ultraclean components.

So, spin coaters, resist and such?

 

[FairAndUnbiased]

So, spin coaters, resist and such?

Overall it's more like ALD precursors, chemical/gas handling, and reactor coatings + associated services. it's a huge operation, I'm only part of it (analytical R&D and custom pilot scale equipment).

Resists are very different, companies that do resists typically are part of more traditional organic/petrochemical operation i.e. DuPont or are more specialized in resist (particular EUV which requires very different skills i.e. Inpria). We're set up for smaller scale materials science and gas/solid phase stuff more than bulk liquids.

My lab in particular makes money through contract R&D projects from the big equipment OEMs, supporting the production facilities for failure analysis and then finally, generating IP for ourselves.

 

[Paul2]

And even with poor demand, you build it, the demand will come, as long as you offer something technically new. Even if it's from the government.

If government doesn't support that, why even care about semiconductor? Just make shoes, ice cream, etc. Just accept slavery.

I just want to communicate the following:

First point, people at state funds have a very long history of spending tremendous amounts of money on vanity projects that tend to close quietly after a few years.

Second point, the state of real part of microelectronics industry, the one that doesn't have the privilege to contend for 12 digit state loans, is far from stellar.

Third point, all of the above, and the last 20 something years rendered it all vividly that keeping doing things under the first point, doesn't improve things under the second point.

Whether it is middle income trap or not, we are not advancing anywhere if the real part of the electronics industry can't stand on its two legs.

This is not a call to throw money on one or another part of the industry. I want people to stop, and think about it. The nation is spending billions a year on something that doesn't work, and doesn't contribute a thing towards this goal.

When anybody here will be passing by Shenzhen, go to SEG plaza, or HQEW building 1 first floor. You will see what China's electronics industry really is.

Overall it's more like ALD precursors, chemical/gas handling, and reactor coatings + associated services. it's a huge operation, I'm only part of it (analytical R&D and custom pilot scale equipment).

Resists are very different, companies that do resists typically are part of more traditional organic/petrochemical operation i.e. DuPont or are more specialized in resist (particular EUV which requires very different skills i.e. Inpria). We're set up for smaller scale materials science and gas/solid phase stuff more than bulk liquids.

My lab in particular makes money through contract R&D projects from the big equipment OEMs, supporting the production facilities for failure analysis and then finally, generating IP for ourselves.

Very glad that there is somebody with whom I can talk about the industry seriously.

I once too had ambitions for process engineering, but in the end things didn't work out.

I tried working in microelectronics instead, an offshoot of Sichuan SOE. They were making a fancy synchronous rectifier for use in "driverless" LED modules. It was a bitter experience, and I am not having any plans to return to it. Plain electronics engineering earn so far much better buck, and if I try really hard, I may open my own engineering practice in the next 6-8 years.