Opinionated - China Chipping Away to Semiconductor Dominance

[oproh]

Great news, more actors like China will make the industry more competitive.
Btw, I didn't know that some technologically advance country like japan is crap when it comes to the semiconductor industry.

 

[Martian2]

Hey @Martian2 ,

Nice report, thanks a lot. I wanna ask you a question regarding SMIC. SMIC offers 40 nm and 28 nm die sizes for it's advanced logic solutions. However 28 nm is a stop gap, and there are limited designs for that die size. 32 nm was the industry standard for that generation. Yet still SMIC does not provide 32 nm solutions.

Do you have any information on when SMIC will start 32 nm technology, or if they will?

Thanks

32nm was a half-node.

28nm has smaller dimensions (including gate pitch) than 32nm.

28nm is superior to 32nm in every way. Semiconductor companies do not go backwards. They utilize older technology for inexpensive chips (such as automotive use). However, SMIC will not build a new factory using an inferior 32nm half-node.

As you can see from the chart below, 28nm is a full node and every semiconductor manufacturer has a plan to move to smaller dimensions. 32nm was a half-node and a stop-gap measure between full nodes. If a company has 28nm, it won't go backwards to build 32nm.

If a company has 32nm, it may decide to skip 28nm and move to 20nm.

I think a 32nm chip only has to be 33% larger in die size to yield the same performance as 28nm. Of course, a 32nm chip would still consume more electricity. The point is that 32nm and 28nm are interchangeable.

Great news, more actors like China will make the industry more competitive.
Btw, I didn't know that some technologically advance country like japan is crap when it comes to the semiconductor industry.

During the 1980s and 1990s, Japan was very strong in semiconductors and electronics. However, the Japanese companies were vertically integrated. This helped the Japanese companies to gain scale very quickly. It is also true in reverse. When Japanese companies started losing market share, they de-leveraged very quickly in both electronics and semiconductors.

I will give you a few examples.

Japanese Sony tvs used to have huge market share. A tv requires a chip to control its functions (e.g. typically called a "driver"). Sony cathode-ray tvs like the Sony Trinitron were made in Japan and kept rising in cost. The market transitioned to LCD displays. The Japanese chip industry suffered for two reasons. LCDs were dominated by Taiwan's AUO and Chi Mei until Korea's LG (which a US federal court ruled had stolen four critical Taiwanese AUO LCD patents) and Samsung came along.

Today, Taiwan is a huge supplier of LCD driver chips for tv and monitor displays. Taiwan ousted Japan's Sony and NEC from the display market.

Similarly, Japan's Epson used to rule the dot-matrix printer market. HP came along and built laser printers. Not only did Epson lose market share, the Japanese dot-matrix chip drivers inside the printers lost their market.

Japan used to own the game console market. Nintendo and Sega used to have Japanese chips inside their game consoles. The market moved to higher performance Xbox AMD-designed chips. Sony's Playstation 4 followed Xbox in using AMD chips. I think Nintendo is abandoning its own custom chips and using AMD chips for its next generation console.

Japan lost all of the chips inside tvs, monitors, printers, and game consoles. Japan has negligible market share in smartphone chips. Japan has basically lost its entire semiconductor and electronics industries. The Taiwanese model of separate companies for component suppliers is superior to the Japanese model of vertical integration.

Taiwan's TSMC fabs run at 90% to 95% efficiency. TSMC aggregates all of the orders from its thousands of customers and maximizes the utility from each plant. Taking Sony as an example, the Japanese plant is either under-utilized or insufficient in capacity when there's a hit product. Sony semiconductor plants fluctuated in capacity utilization based solely on Sony sales in the marketplace.

The Japanese were foolish in not switching to the Taiwanese disaggregated model. In Taiwan, MediaTek will design chips for anybody. The Japanese were limited to designing chips only for their keiretsu. Taiwan's MediaTek can tap the worldwide market. A Japanese competitor was limited to only a few clients within its business group.

It was a no-brainer that Taiwan would trounce Japan in the long run. Taiwan's TSMC currently controls 60% of the worldwide foundry market and has an annual net profit of around $10 billion. Taiwan delivered a TKO (ie. technical knock-out) to Japan.

 

[Pangu]

32nm was a half-node.

28nm has smaller dimensions (including gate pitch) than 32nm.

28nm is superior to 32nm in every way. Semiconductor companies do not go backwards. They utilize older technology for inexpensive chips (such as automotive use). However, SMIC will not build a new factory using an inferior 32nm half-node.

As you can see from the chart below, 28nm is a full node and every semiconductor manufacturer has a plan to move to smaller dimensions. 32nm was a half-node and a stop-gap measure between full nodes. If a company has 28nm, it won't go backwards to build 32nm.

If a company has 32nm, it may decide to skip 28nm and move to 20nm.

I think a 32nm chip only has to be 33% larger in die size to yield the same performance as 28nm. Of course, a 32nm chip would still consume more electricity. The point is that 32nm and 28nm are interchangeable.

During the 1980s and 1990s, Japan was very strong in semiconductors and electronics. However, the Japanese companies were vertically integrated. This helped the Japanese companies to gain scale very quickly. It is also true in reverse. When Japanese companies started losing market share, they de-leveraged very quickly.

I will give you a few examples.

Japanese Sony tvs used to have huge market share. A tv requires a chip to control its functions (e.g. typically called a "driver"). Sony cathode-ray tvs like the Sony Trinitron were made in Japan and kept rising in cost. The market transitioned to LCD displays. The Japanese chip industry suffered for two reasons. LCDs were dominated by Taiwan's AUO and Chi Mei until Korea's LG (which a US federal court ruled had stolen four critical Taiwanese AUO LCD patents) and Samsung came along.

Today, Taiwan is a huge supplier of LCD driver chips for tv and monitor displays. Taiwan ousted Japan's Sony and NEC from the display market.

Similarly, Japan's Epson used to rule the dot-matrix printer market. HP came along and built laser printers. Not only did Epson lose market share, the Japanese dot-matrix chip drivers inside the printers lost their market.

Japan used to own the game console market. Nintendo and Sega used to have Japanese chips inside their game consoles. The market moved to higher performance Xbox AMD-designed chips. Sony's Playstation 4 followed Xbox in using AMD chips. I think Nintendo is abandoning its own custom chips and using AMD chips for its next generation console.

Japan lost all of the chips insides tvs, monitors, printers, and game consoles. Japan has negligible market share in smartphone chips. Japan has basically lost its entire semiconductor and electronics industries. The Taiwanese model of separate companies for component suppliers is superior to the Japanese model of vertical integration.

Taiwan's TSMC fabs run at 90% to 95% efficiency. TSMC aggregates all of the orders from its thousands of customers and maximizes the utility from each plant. Taking Sony as an example, the Japanese plant is either under-utilized or insufficient in capacity when there's a hit product. Sony semiconductor plants fluctuated in capacity utilization based solely on Sony sales in the market place.

The Japanese were foolish in not switching to the Taiwanese disaggregated model. In Taiwan, MediaTek will design chips for anybody. The Japanese were limited to designing chips only for their keiretsu. Taiwan's MediaTek can tap the worldwide market. A Japanese competitor was limited to only a few clients within its business group.

It was a no-brainer that Taiwan would trounce Japan in the long run. Taiwan's TSMC currently controls 60% of the worldwide foundry market and has an annual net profit of around $10 billion. Taiwan delivered a TKO (ie. technical knock-out) to Japan.

Very nice write up, thanks for sharing.

 

[cirr]

Mon Dec 7, 2015 1:11am EST

BRIEF-TSMC plans advanced manufacturing plant in China with net investment under $3 bln

Dec 7 Taiwan Semiconductor Manufacturing Co Ltd

* Says plans to set up 12-inch wafer manufacturing plant, design service centre in Nanjing, China

* Says planned capacity for Nanjing wafer plant is 20,000 12-inch wafers per month

* Says Nanjing plant to begin volume production in second half 2018 of 16-nanometer process technology

* Says net investment in Nanjing plant to be less than $3 billion Source text for Eikon: [ID:bit.ly/PuEXUE] Further company coverage: (Reporting by J.R. Wu)

BRIEF-TSMC plans advanced manufacturing plant in China with net investment under $3 bln| Reuters

 

[Lure]

32nm was a half-node.

28nm has smaller dimensions (including gate pitch) than 32nm.

28nm is superior to 32nm in every way. Semiconductor companies do not go backwards. They utilize older technology for inexpensive chips (such as automotive use). However, SMIC will not build a new factory using an inferior 32nm half-node.

As you can see from the chart below, 28nm is a full node and every semiconductor manufacturer has a plan to move to smaller dimensions. 32nm was a half-node and a stop-gap measure between full nodes. If a company has 28nm, it won't go backwards to build 32nm.

If a company has 32nm, it may decide to skip 28nm and move to 20nm.

I think a 32nm chip only has to be 33% larger in die size to yield the same performance as 28nm. Of course, a 32nm chip would still consume more electricity. The point is that 32nm and 28nm are interchangeable.

Hey thanks for your answer. But there is small issue I wanna remind. Actually 32 nm was that generation's standard node size and 28 nm was the stop gap.

SPIE | Proceeding | Challenges for the 28nm half node: Is the optical shrink dead?

GlobalFoundries: 32nm and 28nm under way, moving away from strict SOI | Chips | Geek.com

IBM, GlobalFoundries move to 28nm process tech- The Inquirer

And this second issue is that of course the smaller the node is it's much more energy efficient and better. But there are a lot of companies out there which made their designs for 32 nm and most of them won't be able to easily adapt 28 nm technology. Of course 28 nm is more advanced, but as you've mentioned there are a lot of IC designers that doesn't work cutting edge. Actually SMIC owes a lot to it's robust 40 nm and 65 nm technology for it's current surge in revenues and profit.

MOQ is another issue for many small companies out there as well. They sometimes work with bigger nodes to have lower MOQ's. Most of the time small companies can't afford large MOQ's for cutting edge nodes.

Here's an interview with SMIC's CEO Tzu-Yin Chiu,

In the midst of SMIC's financial achievements, Chiu last week received the coveted Environment, Health, and Safety (EHS) leadership award from SEMI, the global semiconductor industry association.

With breathing room at last, SMIC now talks about the company's various milestones in its environmental, safety, and philanthropy activities. Chiu touched upon a "SMIC Liver Transplant Program for Children," an initiative launched last year to contribute 2 million RMB to fund liver transplants for impoverished children in China.

During a one-on-one interview with EE Times last week, Chiu didn't hesitate to express SMIC's interest in expansion -- beyond China.

Asked about IBM fabs that may be up for sale, as Big Blue executes its plan to withdraw from the semiconductor business, Chiu said he's interested. "Never rule out the possibility," he said. However, he quickly added, "Of course, we are aware of some of the issues... IBM is, after all, the jewel of the United States."

SMIC is rehabilitated. China's leading foundry's operation is much more stable. The company is building its revenue and profit growth based on, not a wild, but a modest, capacity buildup of 6.7% per year.

But is this all we can expect? Is this as high as SMIC gets?

SMIC's critics worry that SMIC, under Chiu's leadership, might have already given up the dream of directly competing with the world's Tier 1 foundries such as Taiwan Semiconductor Manufacturing Co. (TSMC), Samsung, and Intel.

One long-time semiconductor industry observer based in Shanghai, who spoke on the condition of anonymity, told EE Times, "The gap between SMIC and TSMC is not narrowing, but rather, widening larger in the last few years."

Indeed, while leading fab owners are busy talking about a 14 nm process node, SMIC says its 28 nm process node is "now frozen," allowing potential customers to test and verify SMIC's newest node.

If this is not a concern for SMIC, what other priorities does the company have in mind? How will SMIC compete in the long run?

EE Times recently sat down with SMIC's CEO in its Shanghai headquarters. Here's an edited version of that Q&A.

EE Times: Some in the industry are worried that the technology gap between SMIC and other leading foundries is widening. What's your view?

Tzu-Yin Chiu: Probably not with TSMC... but I think we are narrowing our technology gap with other foundries. We are keeping pace with the industry, and we are quite confident of the progress we're making.

EE Times: What other foundries are you referring to?

Chiu: I'd rather not name names.

EE Times: Help me understand with which specific technologies you think you are narrowing the gap with your competitors.

Chiu: For one, we're very happy with our 40nm/45nm ramp. In 2013, 40nm/45nm revenue contribution was more than $200 million. [SMIC's 45/40 nanometer revenue increased significantly, to account for 16% of the revenue in the second half of 2013.]

Second, our 28nm node technology for both high-k metal gate (HKMG) and POLY/SiON processes were frozen by the end of 2013. Through our Multi-Project Wafer offering, we're entertaining commercial ICs and customer product verification. We are getting very good feedback.


SMIC's technology nodes by percentage

(Source: SMIC)


EE Times: How big is your 28 nm capacity?

Chiu: Beijing will be our main 28 nm fab, where we will have a capacity of 6,000 wafers per month by the end of the third quarter this year. We're moving our equipment to Beijing as we speak. But we also have a capacity in Shanghai, capable of addressing the 28 nm technology.

With Beijing and Shanghai combined, our 28 nm capacity is 15,000 wafers per month.

EE Times: Is that enough?

Chiu: It all depends on the needs of our customers. As you know, our 28 nm process technology is fungible. In other words, those new 28 nm process lines are also capable of 40 nm products. Our plan is that over the next three years, we will build our Beijing facility to have a capacity of 35,000 wafers per month.

EE Times: What do you have to do in order to make that happen?

Chiu: We need to bring in customers and ramp up our technology. [SMIC already did the first Multi-Project Wafer (MPW) late last year. The company is planning on four more MPW shuttles in 2014.]

We also need to make sure that there is enough capital for full ramp-up to 35,000 wafers per month.


SMIC's 28 nm technology milestones

(Source: SMIC)

EE Times: I understand that your Beijing project is 55 percent funded by SMIC and 45 percent by other JV shareholders. You previously said that you're spending about $570 million for your new Beijing project. Is your Beijing fab sufficiently funded?

Chiu: Yes, we're getting what we need, and we're confident of it. This is well within our means.

EE Times: I've been hearing about the Chinese government's strong interest in investing really big money into the domestic semiconductor industry -- over the next five to 10 years -- as part of the nation's initiative to encourage innovation and advance its economy. I'd imagine SMIC could be a big beneficiary of that.

Chiu: We'd welcome the policy to encourage investment in semiconductors. But we are not aware of any details.

EE Times: We all agree about the stability and focus you've brought to SMIC. What were three major factors that you think contributed to the company's steady growth last year?

Chiu: This has been a result of the efforts by the whole team.

But first, I should point out that the successful ramp-up of new technology -- namely, 40nm/45nm processes -- contributed to our revenue in the past year.

Second, our differentiated technology in such areas as CMOS image sensors, power management ICs, and embedded non-volatile memory. All three differentiated technologies combined, we achieved an average of more than 40 percent growth in revenue.

Third, robust growth in China contributed to our success. Forty percent of our revenue comes from Chinese customers.

EE Times: Now that we have begun to hear about the pending merger among Chinese fabless companies, such as the one between Spreadtrum and RDA, are we going to see the number of your Chinese customers decline?

Chiu: We see some Chinese companies are breaking off the pack. We're engaged with all the winners.

EE Times: What about customers outside China?

Chiu: We are going global. We have new engagements all over the world. We have top US fabless companies as our long-term customers over the last 10 years. They're silent, but they're very persistent customers. Among those moving from 40 nm to 28 nm, some are top 10 global fabless companies.

EE Times: How's your CMOS image sensor business going?

Chiu: You know that we have a backside illumination platform for our CMOS image sensors. [BSI is a special way of arranging the imaging elements to increase the amount of light captured, thus improving low-light performance. In essence, it removes the readout circuitry and interconnects from the light path, and illuminates the sensor from the backside.]

With the rise of "selfies," 2 megapixel sensors are in big demand for the front camera of a cellphone. They're in high-volume production now. Our customers' demand for 5 megapixel and 8 megapixel image sensors has grown rapidly.

EE Times: Previously, you talked about the Chinese government's mandate, under which magnetic cards will change to IC cards in 2015. Our understanding is that SMIC developed an embedded EEPROM platform, which had been adopted by a majority of China's bankcard IC design houses. What's the latest?

Chiu: Four out of six UnionPay-qualified bank cards use our platform. Most of our customers are verifying their products, and small-volume production will begin in the second half of this year. A more significant ramp-up and revenue contribution are expected in 2015.

EE Times: How's SMIC's plan for MEMS?

Chiu: SMIC has a very successful program with Silicon Labs on CMEMS [designed to allow direct post-processing of high-quality MEMS layers on top of Silicon Labs' RF/mixed-signal CMOS technology.]

EE Times: But that's mainly for manufacturing CMEMS-based MEMS oscillators. What about other MEMS?

Chiu: Our plan is to closely work with our customers to develop proprietary MEMS process technology. We also have our own capable MEMS teams and a set of MEMS capabilities on our own.

EE Times: What are those?

Chiu: We're not ready to reveal specifics. But our plan is to use our new fab where both CMOS image sensors and MEMS can be manufactured. Our goal is to leverage a part of the special tools required for CMOS image sensor production for MEMS.

EE Times: What new fab?

Chiu: We're actually building a new CMOS Image Sensor [CIS] ecosystem. Our plan is to use this ecosystem for both CIS and MEMS.

First, we already have the 28nm/40nm front-end facility in Shanghai.

Second, we established a joint venture with Toppan. With Toppan, we manufacture on-chip color filters and micro lenses for CMOS image sensors.

Last October, SMIC formed an R&D and manufacturing center dedicated to vision, sensors, and 3D IC, with a mission to consolidate manufacturing capabilities for silicon-based sensors, thru-silicon-via [TSV] technology and other middle-end wafer process [MEWP] technologies.

Then, just last month, we announced a new JV with Jiangsu Changjiang Electronics Technology Co. Ltd., China's largest backend house. The new JV will be responsible for 12-inch bumping and related testing. JCET will also build advanced back-end package production lines nearby.

So, all told, we will be leveraging the network of CIS ecosystem for MEMS as well.

* * *
The SMIC CEO said that SMIC must make sure it meets the needs of the Chinese market and Chinese products. For that matter, in responding to Chinese customers' strong demand for 8-inch displays, SMIC is now bringing up an 8-inch fab in Shenzhen. "We have procured the second-hand equipment, and we're installing them right now," Chiu said. The plan is to build a capacity of 50,000 wafers per month. "Our specialty technology products will be made at the 8-inch fab. They include PMIC and image sensors."

Clearly, SMIC;s focus today is what China needs right now. But what about high-end advanced technology? Should I even ask about SMIC's plan for 14 nm process?

Chiu said, "We will be ready with FinFET at 14 nm process by the end of 2016." He said it as a matter of fact. Somehow, he never made it sound like a big deal.

Will SMIC Narrow Tech Gap? | EE Times

He downplayed the 14 nm node in 2016 which actually marks the drop of technological gap between SMIC and tier 1 fabs to 2 years. However he braggs a lot about 40 nm and 65 nm. If this was my company I'd be popping champaigns celebrating 14 nm technology in 2016. However he is so calm about this.

 

[Martian2]

The money is in 28nm.

TSMC is the foundry industry leader. It derives a full one-third of its revenue from 28nm.

TSMC's revenue is about $26 billion this year. One third of $26 billion is about $8 billion.

28nm is an $8 billion market.

If you're a competitor, you need to take away some of the business from TSMC's 28nm customers.

I don't think TSMC has anything at 32nm. I think 32nm is mostly Samsung's in-house Exynos.
----------

TSMC is the king of 28nm (May 2015 citation)

"TSMC started its commercial 28nm process five years ago and now accounts for over 75 per cent of the global 28nm foundry market. In addition, the 28nm products also contributed 30 per cent to TSMC's total revenues in 2014."
----------

"TSMC Q4 2014 revenue by node"
Source: Intel Or TSMC? Which Deserves Your Investment Dollar? - Intel Corporation (NASDAQ:INTC) | Seeking Alpha

 

[Shotgunner51]

It was a no-brainer that Taiwan would trounce Japan in the long run. Taiwan's TSMC currently controls 60% of the worldwide foundry market and has an annual net profit of around $10 billion. Taiwan delivered a TKO (ie. technical knock-out) to Japan.

Today, Taiwan is a huge supplier of LCD driver chips for tv and monitor displays. Taiwan ousted Japan's Sony and NEC from the display market.

Good job!

 

[Martian2]

TSMC starts mass production of 10nm in 2016. One year ahead of INTEL. | SemiWiki

Daniel Nenni at SemiWiki is a 30-year semiconductor industry veteran. He personally talks to a lot of the important people and companies in the semiconductor industry. According to his latest information, TSMC will beat INTEL to 10nm and 7nm (read his article below).
----------

SemiWiki.com - 20% Growth and 10nm for TSMC in 2016!

 

[cirr]

MediaTek, HLMC announce tape-out of 28nm mobile chips

MediaTek, HLMC announce tape-out of 28nm mobile chips

Jessie Shen, DIGITIMES, Taipei [Monday 16 November 2015]

MediaTek and Shanghai Huali Microelectronics (HLMC) have completed the tape-out of a 28nm mobile chip design, according to the China-based 12-inch foundry.

“This 28nm tape-out with MediaTek marks that HLMC has steadily gained great experiences in advanced process technology management and yield improvement, which has paved a solid basis for the further cooperation between HLMC and MediaTek,” said Jack Qi Shu, VP of sales & marketing at HLMC, in a statement.

In December 2014, HLMC announced it is working closely with MediaTek on 28nm technology development and wafer manufacturing services.

Previous reports quoted industry sources as saying that HLMC and MediaTek had discussed about expanding their partnership to FinFET chips, but none of the companies confirmed the reports.

Founded in 2010, HLMC is majority owned by the Shanghai government. HLMC in 2011 ramped its 12-inch wafer fab, and directly entered 65/55nm production.

MediaTek, HLMC announce tape-out of 28nm mobile chips | SVM

 

[cirr]

Chinese Consortium Makes Rival Bid for Fairchild Semiconductor

By MICHAEL J. de la MERCED

DEC. 8, 2015

SAN FRANCISCO — A trans-Pacific bidding war for Fairchild Semiconductor may be in the offing, as a consortium of Chinese buyers emerged to make a rival offer for the chip maker.

A group led by China Resources Holdings, a state-owned conglomerate, has bid roughly $2.5 billion for Fairchild, according to a person briefed on the matter. That proposal, amounting to $21.70 a share in cash, tops a $20-a-share offer from ON Semiconductor that Fairchild accepted last month.

Also participating in the Chinese takeover effort is Hua Capital Management, an investment firm that led the takeover of the imaging chip maker OmniVision Technologies last spring.

Fairchild acknowledged in a statement on Tuesday that it had received an unsolicited takeover bid that was higher than ON Semiconductor’s offer, though it did not disclose the identity of the suitor. Bloomberg News later reported that the Chinese entities were behind the bid.

The move by the China Resources and Hua highlights the continued interest by Chinese buyers in the chip-manufacturing industry. This year, Tsinghua Unigroup, China’s biggest semiconductor maker, weighed a takeover of Micron Technology, a maker of memory chips.

Adding Fairchild, whose chips help regulate power use in computers and other devices, could bolster the semiconductor arm of China Resources, an enormous conglomerate whose other arms include supermarkets and power plants.

The China Resources consortium emerged as a suitor late in an auction of Fairchild earlier this year and was willing to pay at least $20.20 a share, according to the person briefed on the matter. But by that point, Fairchild essentially had committed to striking a deal with ON Semiconductor.

Fairchild’s agreement with ON Semiconductor does not include a so-called go-shop period that allows the company to solicit higher takeover offers. But the chip maker can evaluate unsolicited bids that it receives.

In its statement on Tuesday, the company said that it would review the new proposal, though its board was still recommending that shareholders accept ON Semiconductor’s deal.

One potential hurdle for the Chinese consortium is the prospect of a tough review by the Committee on Foreign Investment in the United States, an American government panel that reviews takeovers by foreign buyers. That regulator has sometimes looked unfavorably on Chinese companies bidding for American tech companies, though the China Resources group is likely to argue that Fairchild’s chips are not vital to the national security of the United States.

http://www.nytimes.com/2015/12/09/b...val-bid-for-fairchild-semiconductor.html?_r=0

 

[cirr]

China chips away at U.S., Taiwan semiconductor dominance

Dec. 8, 2015, 4:08 PM

By Yimou Lee

HONG KONG (Reuters) - China's multi-billion dollar drive towards self-reliance in semiconductors has spawned a cluster of chip designers that industry experts say could eventually rival leaders Qualcomm Inc and MediaTek Inc .

The world's second-largest economy now boasts nine companies that design and sell chips in the global top 50 from just one in 2009. Clients such as Chinese smartphone manufacturers have also helped compatriot chip designers amass a market share of almost a fifth, according to data analyst TrendForce.

The rise of Chinese designers such as Huawei Technologies Co Ltd [HWT.UL] subsidiary HiSilicon and Spreadtrum Communications comes as the government ploughs funds into home-grown technology to reduce cyber-security risk, following revelations in 2013 of U.S. global cyber-snooping programs.

The revelations have made China a harder place for U.S. tech firms to do business, with Qualcomm saying as recently as last month that it faced delays closing licensing agreements. In contrast, sales at Chinese designers are set to surge this year, some by as much as 40 percent, said researcher IC Insights.

"The Chinese fabless industry is expanding by leaps and bounds," said Bernstein analyst Mark Li, referring to designers which contract out fabrication to so-called foundries such as Taiwan Semiconductor Manufacturing Co Ltd (TSMC) .

Chinese chip designers lag top rivals in terms of technology by four to five years yet have the potential to disrupt the global chip supply chain, industry experts and executives said.

But in terms of size, China is likely to seize second place in the $20 billion-plus chip design industry from Taiwan this year, Li said.

PAST MISTAKES

Chinese factories use over 60 percent of the world's chips annually, and in 2013 imported more chips by value than crude oil. To promote domestic development, the government has tasked chip firms with raising revenue by more than 20 percent annually and building "a group of world-class companies" by 2030.

China's list of chip design hopefuls include HiSilicon plus Spreadtrum and RDA Microelectronics Inc - both controlled by state-backed Tsinghua Unigroup Ltd - as well as All Winner Technology Co Ltd , Leadcore Technology, Galaxycore Microelectronics and Goodix Technology.

"Only by being a market leader can you be profitable," said Tsinghua Unigroup Chairman Zhao Weiguo.

Through a $21.7 billion national fund, as well as at least five other government-led investment vehicles in cities such as Beijing, Shanghai and Nanjing, China has approximately $32 billion under management to build national champions in the chip ecosystem, according to consulting McKinsey & Co.

"Their IC design can become a strong force in a few years," TSMC co-Chief Executive Officer Mark Liu said in a recent interview, referring to chips as integrated circuits (IC).

"However, the system has to reward innovation. You cannot just want market share and dump a lot of low price products into the market. That is not going to help the Chinese IC design sector to grow," Liu said. "So there are good parts and I hope they avoid the bad parts."

There is concern in the industry about a repeat of China's previous efforts to develop industries, such as solar panels and liquid crystal displays (LCD), where overzealous investment led to oversupply and plunging prices.

China made up 14 percent of the global LCD market last year from 3 percent in 2010, while the industry's average profit margin declined to 1.2 percent from 7.8 percent over the same period, wrote Bernstein's Li in a recent report.

"China will not stop until it dominates the market, with value and economics being destroyed every single time," said Li.

(Reporting by Yimou Lee, Miyoung Kim, J.R. Wu; Additional reporting by Paul Carsten; Editing by Christopher Cushing)

China chips away at U.S., Taiwan semiconductor dominance - Business Insider

 

[war&peace]

Good going China

 

[cirr]

UMC to Start 12-inch Fab in China Ahead of Schedule

Alan Patterson

12/7/2015 05:48 PM EST

TAIPEI — United Microelectronics Corp. (UMC), the world’s third-largest foundry, said it expects to open a joint venture 12-inch fab in China about two months earlier than the company expected.

Production will start around the end of the third quarter of 2016, a month or two earlier than UMC’s original expectation for the fourth quarter next year, according to company spokesman Richard Yu. The speed of construction in China has been fast, he said.

UMC is partnering with Fujian Electronics & Information Group and the Xiamen municipal government. The total investment in the fab located in the city of Xiamen on the southeast coast of China is approximately $6.2 billion. UMC will hold about 21% of the venture but will have six out of nine directors on the company board.

The fab will benefit from its location in one of the world’s fastest growing chip markets, which has been chugging along at an annual rate exceeding 10%, according to IC Insights. Although China’s chip market is forecast to be worth $152 billion by 2019, the nation still imports most of its chips.

China’s largest foundry, Semiconductor Manufacturing International Corp. (SMIC), may have been the world’s only chipmaker to run at full utilization during the third quarter this year on strong demand from customers in China.

TSMC’s 12-inch plans

While UMC’s joint venture is the first 12-inch fab in China with investment from a Taiwanese company, UMC’s top rival, Taiwan Semiconductor Manufacturing Co. (TSMC) today announced a plan to build its first wholly owned 12-inch fab in China, slated to start production in the second half of 2018.

The fab plans signal a change for the Taiwanese chipmakers, which have kept most of their production in Taiwan. The island accounts for about a fifth of the world’s semiconductor production.

The Taiwan government restricts semiconductor investments in China on concerns the island will lose jobs and technology to China. Relations between Taiwan and China have been hostile since the Nationalist Party was overthrown militarily by the Communist Party in 1949 and fled to Taiwan in defeat. China considers Taiwan a renegade province and has not ruled out the possibility of taking the island by force.

UMC’s joint venture fab in Xiamen will make chips with geometries no finer than 40nm under Taiwan’s regulations. UMC’s most advanced process technology is at the 28nm mode.

UMC to Start 12-inch Fab in China Ahead of Schedule | EE Times

 

[oproh]

Great to see China's share getting bigger in this important industry.

 

[cirr]

Commentary: TSMC new fab to boost China semiconductor industry

Cage Chao, Taipei; Steve Shen

DIGITIMES [Thursday 10 December 2015]

The announcement by Taiwan Semiconductor Manufacturing Company (TSMC) to build a 12-inch wafer manufacturing facility in Nanjing, China heralds the establishment of a cluster of semiconductor industry in China as well as a takeoff of China's semiconductor industry in coming years.

Although the planned capacity of 20,000 12-inch wafers a month for the new facilities will account for only a 2.5% of TSMC's total production capacity, it comes as a boost to China's ambition to raise its self-sufficiency in chip production. China has said that it wants to ramp up the self-sufficiency rate for production of chips to 70% by 2025.

TSMC chairman Morris Chang has pointed out that in view of the rapid growth of China's semiconductor market, TSMC has to establish a 12-inch wafer fab and a design service center in China to provide closer support to its customers there and to further expand business opportunities.

Revenues generated from the China market accounted for 6% TSMC's total sales in 2014 and are expected to increase to 8% in 2015, making China the third largest market for TSMC trailing after only the US and Taiwan.

Increasing presence of other semiconductor firms in China is another reason for TSMC moving to China. Intel has tied up with Spreadtrum Communications to produce NAND flash chips in China, while Samsung Electronics and SK Hynix already have their own fabs in China producing NAND flash and DRAM chips, respectively.

Fellow Taiwan-based companies United Microelectronics Corporation (UMC) and Powerchip Technology are already moving forward with their plans to cooperate with local partners to build 12-inch wafer fabs in China.

But the competitiveness is actually not an issue for TSMC as its planned facilities in Nanjing is scheduled to begin volume production of 16nm process technology in the second half of 2018, while UMC and Powerchip are not allowed by Taiwan's laws to invest in their most-advance process technologies at their 12-inch JV fabs in China.

A more significant outcome that will come with TSMC's 12-inch fab in Nanjing is that upstream and downstream partners of TSMC, including equipment and semiconductor material suppliers and IC backend service companies, will also flock into Nanjing to form a large-scale semiconductor cluster there, duplicating the successful establishments of the science parks in Taiwan and bringing a boom to China's semiconductor industry.

Commentary: TSMC new fab to boost China semiconductor industry