China chip industry resilient, Intel rival to emerge in 3-5 years?

[Daniel Nenni]

China is surprisingly resilient in the chip war. Credit: DIGITIMES

With China completing its 2022 National Party Congress on March 11, some information released during and before the event raised a few eyebrows.
One example is the comment on local Computer Processing Unit (CPU) manufacturers' competitiveness by an Intel senior executive in China. Rui Wang, SVP of Intel Corporation and chair of Intel China said that she expects to see mainland Chinese enterprises in the CPU space become Intel's competitors in the next 3-5 years, during an interview with local media Guancha.cn.

"So far there has not been any local companies that are able to deal a substantial threat to Intel," said Wang. "But in 3-5 years, it will become clear that local companies will emerge as strong rivals." China contributes more than 25% of Intel's revenues.

At present, local CPUs in China can be divided into three camps: Shanghai Zhaoxin Semiconductor and Hygon, which adopt the X86 core licensing model; Huawei's HiSilicon Kunpeng CPU and Phytium Technology in Tianjin, which adopt the ARM instruction set architecture licensing and self-designed CPU cores; and Loongson and Sunway Microelectronic, which develop their own instruction set architecture and IP cores. These three routes show a low to a high degree of independent control, but the ecological maturity is just the opposite, said Guancha.cn.

Wang was confident that Intel's sound Wintel ecosystem is not easy to beat. Although she hopes there will be local companies to compete with Intel in China, "Intel won't be polite, and will exert its power to compete fairly."

Sounds familiar? IBM was confident that its servers will remain the winner in China, but Inspur, which has been receiving generous investments and subsidies from the Chinese government since 2009, announced in 2014 that it will take over IBM's business in China, replacing IBM servers used by local banks.

"Cyber security" issues were mentioned when local banks sought to replace foreign brand servers with local ones, as soon as the local firms start to master the technologies. Although semiconductor technologies cannot be copied with reverse engineering easily, the "business ecosystem" built over the years with another foreign strategic partner really isn't as invincible as some people would have believed.

Buy local​

Despite the fact that many Chinese companies were put under the Entity List by the US government, somehow those fabless companies still managed to find local foundries to produce their chips with mature processing nodes. According to a press release on December 29, 2021, China Greatwall Technology Group, which holds 31.5% of shares in Phytium Technology, said Phytium has a comprehensive layout in the CPU market, with products covering servers, desktops, and embedded. It claimed that Phytium maintains a high rate of generation upgrades, with five new high-performance CPUs planned for release in 2021 and 2022.

Is China preparing for decoupling or bifurcation? According to China's official media Global Times, the US's sanction reinforced the need to be self-sustaining and resilient, leading China to accelerate domestic replacements in both hardware and software. It claimed some are designed by China with open-source development tools. Meanwhile, the measure by the Chinese government requiring local manufacturers to prioritize procurement from local-brand chips has boosted sales of those local chips by folds.

In a press conference held on February 28, right before the National Party Congress, Xiao Yaqing, China's Minister of Information and Technology, said the chip industry in China had grown 33.3% in 2021 from a year ago, thanks to strong demand for chips due to digital transformation. The spokesman of the Ministry of Information and Technology claimed that 55.3% of corporations with annual sales of more than CNY 20 million in China have managed to complete the digital transformation of critical manufacturing processes, while the penetration of digital R&D tools has reached 74.7%. There are more than 2,000 5G industrial internet projects throughout China, according to the Ministry spokesman.

The chip shortage problem which has impacted auto industries worldwide didn't seem to bother electrical vehicle makers in China, either. A total of 452,000 EVs were produced and 431,000 sold in China in 2021, registering a 130% and 140% growth year-on-year, respectively.
Interestingly, vice Minister Xin Guobin mentioned that they saw a misallocation of chips in the auto industry in China. "Some companies got plenty of chips, but consumers did not want to buy their products, while other auto companies which make popular products were forced to stop production due to chip crunch, resulting in a mismatch of resource and demand," said Xin.

Another report posted by China Semiconductor Industry Association (CSIA) pointed out that the US Entity List and Foreign Direct Product Rule has taken a toll on TSMC's revenue in China, which declined by nearly 30%, to CNY 164.55 billion. If judged by net revenue contribution, the slide is even larger.
According to the 4Q21 management report of TSMC, China contributed 10% of its net revenues in 2021, a 41% drop from 17% in 2020. Eric Chen, DIGITIMES semiconductor analyst explained that Huawei was the second largest contributor to TSMC in 2020, but TSMC stopped supplying chips to Huawei in 2021 due to FDPR, hence the steep drop.

It was strong demand elsewhere that made it up to TSMC 's sales lost due to FDPR. It is noteworthy that the CSIA report mentioned that TSMC enjoyed a 20.37% increase in sales to its largest customer, which is likely to be Apple. The contribution of the US and Asia Pacific markets to its net revenues increased by 3 percentage points each while EMEA increased by 1 percentage point in 2021, compared to a year ago.

China chip industry resilient, Intel rival to emerge in 3-5 years?

With China completing its 2022 National Party Congress on March 11, some information released during and before the event raised a few eyebrows.

www.digitimes.com

 

[benb]

It's fascinating to see China incubate and mature these new businesses along the path to competitiveness. Many narrative-busting lessons to be learned here.
China is not beset with huge entrenched monopolies. There used to be a "true north" of semiconductor development, the ITRS, now China's 5 year plans fill the void. And why not? The US abandoned the institution that brought consistent huge progress and competition. China is playing a humble role here in seeing the need and filling it.

 

[coldsolder215]

The Rus/Ukr conflict and resulting economic war will cause China to double down. Recently bought a 4 channel 200MHz Siglent scope for $740, the closest comparable thing Tektronix offers is >$2k. Now if they could simply smash the American/German EDA cartel we'd really be cooking with gas. I've had some exposure to Empyrean and it's still an inferior clone, but also Cadence hasn't updated the UI since the 90s so there's gotta be some room for improvement.

 

[Paul2]

The Rus/Ukr conflict and resulting economic war will cause China to double down. Recently bought a 4 channel 200MHz Siglent scope for $740, the closest comparable thing Tektronix offers is >$2k. Now if they could simply smash the American/German EDA cartel we'd really be cooking with gas. I've had some exposure to Empyrean and it's still an inferior clone, but also Cadence hasn't updated the UI since the 90s so there's gotta be some room for improvement.

De-Cadence is a horror show, and a liability for the entire industry. FPGA toolchain software is an equivalently bad proprietary mess.

Open source silicon cannot emerge as an equivalent to Linux for hardware industry unless these two things above hold the whole industry hostage.

It's really a case of national security crisis now for US to back off on religious like blind trust in "intellectual property" ideology.

 

[Fred Stein]

Fascinating to hear China's focus. Also to hear their success producing EVs. But Tesla produced twice as many in 2021 and they aren't the only US EV car maker.

More curious and confusing is why is this Intel v. China? TMSC and its partners Apple, Qualcomm, NVIDIA, seem to be leading in mfg and design. Also Samsung doing both.

Wang seems to assert that WinTel is not easy to beat. Again, likely true w.r.t. China but not AMD and their foundries.

 

[Jozo035]

...Recently bought a 4 channel 200MHz Siglent scope for $740, the closest comparable thing Tektronix offers is >$2k. ...

Well, they are more expensive for good freason .

I can connect 4 cheap oscilloscopes to signal source with equal splitter and each of them will show different result.

Which is terrible because You spend hours troubleshooting problems caused by Your equipment.

It is bit off topic, but i think it applies in other areas too...

 

[tooLongInEDA]

The Rus/Ukr conflict and resulting economic war will cause China to double down. Recently bought a 4 channel 200MHz Siglent scope for $740, the closest comparable thing Tektronix offers is >$2k. Now if they could simply smash the American/German EDA cartel we'd really be cooking with gas. I've had some exposure to Empyrean and it's still an inferior clone, but also Cadence hasn't updated the UI since the 90s so there's gotta be some room for improvement.

I don't understand this comment. "A cartel is an organization created from a formal agreement between a group of producers of a good or service to regulate supply in order to regulate or manipulate prices."

In 30 years working in and around EDA I've never seen evidence of this. If you think there is, you should report it and perhaps provide some evidence. This is not a word to be carelessly thrown around without evidence - always a chance you'd get sued.

There have always been EDA start-ups and competition. Where EDA companies get fat and lazy, they eventually lose market share. If you think you can do better, start an EDA company and prove it.

The fact that EDA is a market with a small number of large suppliers (like Airbus/Boing for airliners or GE/Rolls Royce for airline engines) is no proof of a cartel. Some industries have very high barriers to entry and high cost/risk of switching. That's just how it is.

 

[benb]

If you assume that in China, there will be an x86 clone chip, at about 10x lower price, and it seems like this always happens, it can lead to questions about reverse engineering, IP theft, and business practices in China. Something unfair going on.
Or you could look at it from China’s perspective. They probably just see a bunch of obstacles in their way on the path to creating a clone 8th or 10th Gen Core chip, and they want to remove those obstacles. So they need their own fabs, their own EDA software, all of it.

I watched a Nebula documentary about Zilog Z80, a better chip than Intel 8008 but it lost out in the end. It highlights how being the best isn’t really ever decisive.

 

[coldsolder215]

I don't understand this comment. "A cartel is an organization created from a formal agreement between a group of producers of a good or service to regulate supply in order to regulate or manipulate prices."

I see that Cadence dominates analog/design entry, Synopsys dominates digital flows, and Mentor dominates layout/backend stuff, but they all offer token competitive options in each others' domains to provide cover for what amounts to a coordinated monopoly. I'm not judging it's an impressive feat of modern capitalism.

In 30 years working in and around EDA I've never seen evidence of this. If you think there is, you should report it and perhaps provide some evidence. This is not a word to be carelessly thrown around without evidence - always a chance you'd get sued.

Hard to see the forest when you're in the trees but it's not hard to dig up discourse on this issue, e.g. https://www10.edacafe.com/blogs/wha...mentor-cartel-canceled-gary-smith-vindicated/

There have always been EDA start-ups and competition. Where EDA companies get fat and lazy, they eventually lose market share. If you think you can do better, start an EDA company and prove it.

Yes until one of the big 3 buys it up and snuffs it out, that is the dream.

 

[coldsolder215]

Well, they are more expensive for good freason .

I can connect 4 cheap oscilloscopes to signal source with equal splitter and each of them will show different result.

I think this trope has sailed, how else could China build a 5G network that is 10x that of USA? I've seen screenshots of a 500MHz waveform on their 200MHz 2MS/s offering but maybe that's poor QC and it goes the other way too. They laughed when Mao had people smelting pig iron in their backyards 60 years ago, and now China does more steel better than anyone else.

 

[tooLongInEDA]

I see that Cadence dominates analog/design entry, Synopsys dominates digital flows, and Mentor dominates layout/backend stuff, but they all offer token competitive options in each others' domains to provide cover for what amounts to a coordinated monopoly. I'm not judging it's an impressive feat of modern capitalism.

Hard to see the forest when you're in the trees but it's not hard to dig up discourse on this issue, e.g. https://www10.edacafe.com/blogs/wha...mentor-cartel-canceled-gary-smith-vindicated/

Yes until one of the big 3 buys it up and snuffs it out, that is the dream.

Where do I start with this ?

Synopsys does not dominate digital design flows. Cadence is also very strong here. In fact, they've become very strong in layour (place and route) and re-established the position they used to have 30 years ago. Partly through innovation and sheer hard work over many years and partly through unforced errors by Synopsys.

Mentor (it's Siemens these days) has almost nothing in layout. Hasn't been a significant player in layout for over 20 years.

I've worked for 2 of the big 3 EDA companies and never seen any evidence of a "coordinated monopoly". There is plenty of competition and customers are reluctant to allow monopolies to develop.

Remember also that creating a near monopoly market share through developing products that customers prefer is not illegal, not necessarily "immoral". It's often just doing a good job while others perhaps don't. Do we want to punish success ?

Indeed, there have been start-ups that broke through - Avanti (based on stolen Cadence IP) and Magma to name just two. But it is very hard to do that now - the code base is just too large to create everything from scratch in any realistic timeframe and the cost and technical challenge too large to get anty realistic ROI. That's just the nature of this business. And nothing to do with any so-called "cartel".

Cadence has spent the last 30 years fighting to get a competitive logic synthesis capability. They've finally achieved some success at the third (or possibly fourth) attempt. Why did they spend all that time and effort if EDA was a cartel ? It's highly unlikely they've made much/any money from these efforts yet. But it's a strategic imperative to have a full digital design flow these days.

I also worked for a startup. Got acquired by a large EDA company. They used the technology and kept the people. It was not "snuffed out".

 

[hist78]

China is surprisingly resilient in the chip war. Credit: DIGITIMES

With China completing its 2022 National Party Congress on March 11, some information released during and before the event raised a few eyebrows.
One example is the comment on local Computer Processing Unit (CPU) manufacturers' competitiveness by an Intel senior executive in China. Rui Wang, SVP of Intel Corporation and chair of Intel China said that she expects to see mainland Chinese enterprises in the CPU space become Intel's competitors in the next 3-5 years, during an interview with local media Guancha.cn.

"So far there has not been any local companies that are able to deal a substantial threat to Intel," said Wang. "But in 3-5 years, it will become clear that local companies will emerge as strong rivals." China contributes more than 25% of Intel's revenues.

At present, local CPUs in China can be divided into three camps: Shanghai Zhaoxin Semiconductor and Hygon, which adopt the X86 core licensing model; Huawei's HiSilicon Kunpeng CPU and Phytium Technology in Tianjin, which adopt the ARM instruction set architecture licensing and self-designed CPU cores; and Loongson and Sunway Microelectronic, which develop their own instruction set architecture and IP cores. These three routes show a low to a high degree of independent control, but the ecological maturity is just the opposite, said Guancha.cn.

Wang was confident that Intel's sound Wintel ecosystem is not easy to beat. Although she hopes there will be local companies to compete with Intel in China, "Intel won't be polite, and will exert its power to compete fairly."

Sounds familiar? IBM was confident that its servers will remain the winner in China, but Inspur, which has been receiving generous investments and subsidies from the Chinese government since 2009, announced in 2014 that it will take over IBM's business in China, replacing IBM servers used by local banks.

"Cyber security" issues were mentioned when local banks sought to replace foreign brand servers with local ones, as soon as the local firms start to master the technologies. Although semiconductor technologies cannot be copied with reverse engineering easily, the "business ecosystem" built over the years with another foreign strategic partner really isn't as invincible as some people would have believed.

Buy local​

Despite the fact that many Chinese companies were put under the Entity List by the US government, somehow those fabless companies still managed to find local foundries to produce their chips with mature processing nodes. According to a press release on December 29, 2021, China Greatwall Technology Group, which holds 31.5% of shares in Phytium Technology, said Phytium has a comprehensive layout in the CPU market, with products covering servers, desktops, and embedded. It claimed that Phytium maintains a high rate of generation upgrades, with five new high-performance CPUs planned for release in 2021 and 2022.

Is China preparing for decoupling or bifurcation? According to China's official media Global Times, the US's sanction reinforced the need to be self-sustaining and resilient, leading China to accelerate domestic replacements in both hardware and software. It claimed some are designed by China with open-source development tools. Meanwhile, the measure by the Chinese government requiring local manufacturers to prioritize procurement from local-brand chips has boosted sales of those local chips by folds.

In a press conference held on February 28, right before the National Party Congress, Xiao Yaqing, China's Minister of Information and Technology, said the chip industry in China had grown 33.3% in 2021 from a year ago, thanks to strong demand for chips due to digital transformation. The spokesman of the Ministry of Information and Technology claimed that 55.3% of corporations with annual sales of more than CNY 20 million in China have managed to complete the digital transformation of critical manufacturing processes, while the penetration of digital R&D tools has reached 74.7%. There are more than 2,000 5G industrial internet projects throughout China, according to the Ministry spokesman.

The chip shortage problem which has impacted auto industries worldwide didn't seem to bother electrical vehicle makers in China, either. A total of 452,000 EVs were produced and 431,000 sold in China in 2021, registering a 130% and 140% growth year-on-year, respectively.
Interestingly, vice Minister Xin Guobin mentioned that they saw a misallocation of chips in the auto industry in China. "Some companies got plenty of chips, but consumers did not want to buy their products, while other auto companies which make popular products were forced to stop production due to chip crunch, resulting in a mismatch of resource and demand," said Xin.

Another report posted by China Semiconductor Industry Association (CSIA) pointed out that the US Entity List and Foreign Direct Product Rule has taken a toll on TSMC's revenue in China, which declined by nearly 30%, to CNY 164.55 billion. If judged by net revenue contribution, the slide is even larger.
According to the 4Q21 management report of TSMC, China contributed 10% of its net revenues in 2021, a 41% drop from 17% in 2020. Eric Chen, DIGITIMES semiconductor analyst explained that Huawei was the second largest contributor to TSMC in 2020, but TSMC stopped supplying chips to Huawei in 2021 due to FDPR, hence the steep drop.

It was strong demand elsewhere that made it up to TSMC 's sales lost due to FDPR. It is noteworthy that the CSIA report mentioned that TSMC enjoyed a 20.37% increase in sales to its largest customer, which is likely to be Apple. The contribution of the US and Asia Pacific markets to its net revenues increased by 3 percentage points each while EMEA increased by 1 percentage point in 2021, compared to a year ago.

China chip industry resilient, Intel rival to emerge in 3-5 years?

With China completing its 2022 National Party Congress on March 11, some information released during and before the event raised a few eyebrows.

www.digitimes.com

The person gave this assessment is Rui Wang, Senior Vice President of Intel China. Should we consider the possibility that this is an Intel's effort to convince US government more subsidies are needed?

 

[michele.e.reitz]

oreign Direct Product Rule

Where do I start with this ?

Synopsys does not dominate digital design flows. Cadence is also very strong here. In fact, they've become very strong in layour (place and route) and re-established the position they used to have 30 years ago. Partly through innovation and sheer hard work over many years and partly through unforced errors by Synopsys.

Mentor (it's Siemens these days) has almost nothing in layout. Hasn't been a significant player in layout for over 20 years.

I've worked for 2 of the big 3 EDA companies and never seen any evidence of a "coordinated monopoly". There is plenty of competition and customers are reluctant to allow monopolies to develop.

Remember also that creating a near monopoly market share through developing products that customers prefer is not illegal, not necessarily "immoral". It's often just doing a good job while others perhaps don't. Do we want to punish success ?

Indeed, there have been start-ups that broke through - Avanti (based on stolen Cadence IP) and Magma to name just two. But it is very hard to do that now - the code base is just too large to create everything from scratch in any realistic timeframe and the cost and technical challenge too large to get anty realistic ROI. That's just the nature of this business. And nothing to do with any so-called "cartel".

Cadence has spent the last 30 years fighting to get a competitive logic synthesis capability. They've finally achieved some success at the third (or possibly fourth) attempt. Why did they spend all that time and effort if EDA was a cartel ? It's highly unlikely they've made much/any money from these efforts yet. But it's a strategic imperative to have a full digital design flow these days.

I also worked for a startup. Got acquired by a large EDA company. They used the technology and kept the people. It was not "snuffed out".

I would have to agree by many of the points above, also been too long around EDA. China will always look to MAKE or BUY any technology they need, and they continue to ask technology analysts "how do I spend these $B? Should we make EDA tools? should we design FPGAs? should we build our own x86 or other cores?" My answer was always that they should look deep into the man-years needed to compete. In hardware I think it may be easier than in EDA sw. However, the startup EDA companies can ignore all the legacy approaches and really re-define the landscape. I've worked for all of the EDA companies (or acquired companies) and 2 startups and have been watching this for over 25+ years--is going to be interesting.

 

[tonyget]

Interesting video Russians discussing the possibilty of making a completely home-grown smartphone with Russian made components after this sanction

The video is in Russian language，you can turn on the subtitle and choose english translation

 

[hist78]

Interesting video Russians discussing the possibilty of making a completely home-grown smartphone with Russian made components after this sanction

The video is in Russian language，you can turn on the subtitle and choose english translation

Vyacheslav Martynov who was the CEO of Yota Devices (produced the first Russian smartphone). He said it's impossible to make a Russian smartphone without international suppliers, technology, market, and ecosystem.

He even went further to use South and North Korea as an example to illustrate how difficult it is. He indirectly hinted that after Qualcomm, Mediatek, TSMC, Samsung, etc. stopped doing business with Russia, there isn't much hope to have a Russian produced smartphone.

 

[tonyget]

Vyacheslav Martynov who was the CEO of Yota Devices (produced the first Russian smartphone). He said it's impossible to make a Russian smartphone without international suppliers, technology, market, and ecosystem.

He even went further to use South and North Korea as an example to illustrate how difficult it is. He indirectly hinted that after Qualcomm, Mediatek, TSMC, Samsung, etc. stopped doing business with Russia, there isn't much hope to have a Russian produced smartphone.

There are many Russians say that Yota phone was nothing but a scam project，the company took lots of fund from Russian government to develop an indigenous phone. But what they did was just outsource the production to Chinese OEM then stick their logo on the phone，there isn't much if any Russian content in the phone.

 

[hist78]

There are many Russians say that Yota phone was nothing but a scam project，the company took lots of fund from Russian government to develop an indigenous phone. But what they did was just outsource the production to Chinese OEM then stick their logo on the phone，there isn't much if any Russian content in the phone.

Yota declared bankruptcy in 2019. Putin, oil, gas, and corruption brought Russia to no where.

 

[smeyer0028]

The current monopoly's roots go back to an attempt in the 1990s
to form a small EDA company organization. The big three
(maybe one of them) sued. The judge ruled that excluding the
big 3 was illegal so they took over the organization.
It meant no cooperation among small
EDA companies. It is impossible to get VC capital in the
EDA market because the answer will be the only way we
can make a profit from our investment is if one of the big
3 buys the company.

There is a need for innovation in the EDA area. For example
Spice goes back to maybe even the late 1960s. Model's
are astoundingly complicated. There has been progress
in the applied mathematics of E&M since then.

 

[tooLongInEDA]

The current monopoly's roots go back to an attempt in the 1990s
to form a small EDA company organization. The big three
(maybe one of them) sued. The judge ruled that excluding the
big 3 was illegal so they took over the organization.
It meant no cooperation among small
EDA companies. It is impossible to get VC capital in the
EDA market because the answer will be the only way we
can make a profit from our investment is if one of the big
3 buys the company.

There is a need for innovation in the EDA area. For example
Spice goes back to maybe even the late 1960s. Model's
are astoundingly complicated. There has been progress
in the applied mathematics of E&M since then.

Why does acquisition by one of the "big 3" not count ? A buyout is an equally valid VC exit as an IPO and does get a return for the VCs. Some large EDA startup buyout valuations still happening.

Once again, "monopoly" is a term with legal meaning (like "cartel") that we should not be throwing around casually. I've never come across a 3 player monopoly.

I've started a separate thread on EDA monopolies and cartels so we can keep the discussion neatly in one place as I expect there are more views to be aired here.

 

[VCT]

Interesting video Russians discussing the possibilty of making a completely home-grown smartphone with Russian made components after this sanction

The video is in Russian language，you can turn on the subtitle and choose english translation

Russian can make crappy smartphone but people won't buy them. Used iPhone 6 will sell very well in Russia.