China: Fab Boom or Bust?

China’s semiconductor industry continues to expand at a frenetic pace. At present there are nearly two dozen new fab projects in China.

Whether all these fab projects get off the ground is not entirely clear because the dynamics in China remain fluid. What is clear is the motivation behind this building frenzy—China is trying to reduce its huge trade imbalance in ICs. The country continues to import a large percentage of its chips from foreign vendors.

The Chinese government wants to produce more chips within China, and it also wants to keep closer tabs on those ICs for security reasons. As part of the plan, China has lured several multinational chipmakers to build new fabs inside its borders. For multinational chipmakers, the attraction is the ability to get closer to an enormous customer base. GlobalFoundries, Intel, Samsung, SK Hynix, TSMC and UMC all are building new fabs or expanding their existing plants in China.

In addition, China’s domestic chipmakers also are building new fabs. Estimates are all over the map, but there are somewhere between 14 to 22 new fab projects in China today, many of which are under construction or on the drawing board. These projects include both domestic and multinational chipmakers.

It’s doubtful that all fab projects will come to fruition, however. For one thing, some of China’s chipmakers lack the technology to fulfill their fab ambitions. In fact, some plans already have been scrapped.

Plus, China’s chipmakers have announced various fab projects in the past, but many were never built. Based on these and other trends, some analysts estimate that 50% of the new fab projects could fail in China, although others see a higher success rate.

“I don’t want to give a number like 50%, but I wouldn’t want to bet that every new fab will be up and running in China,” said Samuel Wang, an analyst with Gartner. “Everyone is aggressively going forward with all of the projects in China. But until the fabs are actually up and running, anything can happen.”

If all fabs move into production, China’s total 300mm-equalvent output is projected to more than triple, from 400,000 wafer starts per month (wspm) today, to 1.4 million wspm by 2020, according to Gartner. Of that figure, 60% of the output is geared for memory, while 40% is logic, according to the firm.

Even if China doesn’t build all of these fabs, there are still some worrisome signs. For example, a large percentage of the projected capacity from China’s foundry vendors is slated for 28nm processes and above. So over time, some see a potential capacity glut, if not a price war, looming for these technologies. “The price competition could get really strong from 2018 on,” Wang said. “That is the worry in the foundry business.”

Going forward, the challenge for China’s own chipmakers is to move beyond the trailing-edge and develop more advanced technologies. So don’t expect new and leading-edge logic capacity coming from China anytime soon. The exception to the rule is TSMC, which will produce 16nm finFETs in a new China fab by 2018.

China’s fab initiatives
China’s IC efforts are fueling a wave of fab activity in the nation, creating a boom in the equipment industry. In total, China’s fab equipment spending is projected to jump from $6.7 billion in 2017 to $10 billion in 2018, according to SEMI.

Today, there are about 20 200mm fabs and 10 300mm facilities in operation within China, according to estimates from Shanghai Huali, a Chinese foundry vendor. Currently, China is building 7 new 200mm plants and 12 300mm facilities, according to the company.

Each fab has a different production schedule, although the big fab ramp is expected to start in 2018. “We think 2017 spending for wafer fab equipment in China will be similar to last year,” said Arthur Sherman, vice president of marketing and business development at Applied Materials. “Our expectations are for a significant ramp in spending in 2018 and beyond. Based on public announcements and talking to manufacturers, we expect large investments supporting long-term strategies to take place in China over the next several years.”

Others agreed. “(The) majority of these new projects are focusing on trailing edge, high-volume manufacturing of foundry logic devices, and new memory-centric projects emerging in the near future,” said Francis Jen, general manager of sales at KLA-Tencor China. “Legacy/mature node expansion (in) China will also create opportunities for refurbished or even newly-built legacy equipment.”

Still, fab equipment vendors must proceed with caution in China as well. If all of these fabs go into production, China will require at least a whopping $70 billion worth of IC equipment alone by 2020, according to Gartner. But it’s unlikely that all fabs will appear by then. So, the challenge for fab tool vendors is to make the right forecast in China. Otherwise, equipment vendors could end up with a shortfall of gear for the China market or a plethora of unwanted tools in inventory.

Nevertheless, China is the hot spot in the market. Today, the nation produces a large percentage of the world’s electronic products. As a result, China is the world’s largest consumer of chips. These ICs are used in systems, which are exported or resold into China.

China also boasts a sizeable domestic semiconductor industry. In fact, starting in the 1980s, China introduced several initiatives to modernize its IC industry. And over the years, China has launched several domestic foundry vendors and a plethora of fabless design houses. Originally, domestic foundry vendors were supposed to produce a larger share of chips within China and present some new competition in the marketplace.

China has made some progress at each turn, but every initiative has fallen short of expectations. For example, the nation continues to import more chips than it produces by a wide margin. It also lags behind its foreign rivals in IC technology.

To solve the problem, China in 2014 launched an initiative called the “National Guideline for Development of the IC Industry.” At the time, China also created a $19.3 billion fund, which would be used to invest in its domestic IC firms. And over the next decade, it could spend $100 billion across China’s IC sector.

Generally, with this and other initiatives, China hopes to achieve the following goals:

• Produce 40% of its own chips by 2020 and 70% by 2025.
• Accelerate its efforts in 14nm finFETs, advanced packaging, MEMS and memory.
• Acquire technology from a foreign country to fill a gap if the nation can’t develop that technology itself.

The strategy makes sense, at least in theory. “China is becoming a leading consumer of electronics, but they’re not yet producing much of the semiconductor components for those electronics systems,” said David Fried, chief technology officer at Coventor. “It just makes sense for them to bring that semiconductor production inside to satisfy consumer demand. So they’re going to invest a lot in logic and memory. And they are going to try to do it fast.”

So far, though, China’s initiatives in ICs have been a mixed bag. For instance, China-based foundries captured a modest 9.2% of the worldwide pure-play foundry market in 2016, up from 8.5% in 2011, according to IC Insights. “There have been some successes in China,” said Bill McClean, president of IC Insights. “Do I think China’s approach and goals were realized in the foundry business? In general, no.”

Plus, China continues to rely on foreign IC suppliers. In 2016, the nation consumed $112 billion, or 38% of the world’s ICs, according to IC Insights. But chipmakers in China produced only $13 billion worth of chips within China, according to the firm. That represents only 11.6% share of the China market in 2016, up from a mere 9.8% in 2011, the firm said.

Based on those numbers alone, it’s doubtful that China can produce 40% of its own chips by 2020. “(The 40% goal) will not come close to being met,” McClean said.

Fig. 1: Semiconductor manufacturing in China. Source: IC Insights

Still, the country is making progress. “When China last announced a large build-up of fabs, around the year 2000, the GDP of China at that time was around $1 trillion U.S. dollars. And in 2000, China had a zero base in terms of advanced fabs, design houses and semiconductor talent,” observed Linyong “Leo” Pang, chief product officer and executive vice president at D2S. “In 2015, however, China’s GDP was $11 trillion U.S. dollars, more than 10X what it was 15 years earlier. Now, there are over 10 advanced fabs, thousands of design houses, and millions of talented people working in China’s semiconductor industry.”

Over the last decade, Pang also has seen a new crop of engineers entering the semiconductor field in China. “(Many are) moving into the executive level of these new foundries. They are well educated, well trained and very creative,” he said. “Given that, I think this time most of these 20 fab projects, if not all or more, will happen.”

Behind the fab frenzy
But how many of these fabs will be built, when they will be built, and what they will produce is complicated.

There are two categories that need to be considered—memory and foundry/logic. In memory, the nation is expanding on two fronts. On one front, three multinational companies—Intel, Samsung and SK Hynix—are expanding their memory production in fabs within China. Then, as reported in Semiconductor Engineering, several domestic memory makers have emerged. (See http://semiengineering.com/china-unveils-memory-plans/ ) One government-backed company, Yangtze River Storage Technology (YRST), recently announced two multi-fab projects at a staggering cost of $54 billion. YRST hopes to make 3D NAND and DRAM.

Regarding China’s memory efforts, analysts are skeptical. It’s unclear where domestic vendors will obtain the technology. “The new memory makers from China will need more technology support,” said Alex Yang, an analyst with the Market Intelligence & Consulting Institute (MIC), a market research firm in Taiwan. “They may require technology transfers (from foreign companies) to enhance their process capabilities. It won’t be easy for them in the short term.”

Fig. 2: New foundry fabs in China. Source: MIC

Meanwhile, like the memory segment, China’s foundry sector can be sub-divided into two groups—domestic and multinational. The multinationals will likely succeed in building their proposed fabs in China. Time will tell if the domestic foundries will make good on their fab plans.

In both cases, most of the new foundry capacity is geared toward more mature technologies. “We see new fabs coming on line as well as a lot of capacity being added to existing fabs,” said Doug Lawson, executive vice president of corporate marketing and strategy at Axcelis. “This is driven primarily by investments in mature technologies. The is driving a significant amount of this activity, and we expect the momentum to continue. While most of the investment in 2017 is in IoT and mature process technology, 2018 will see the start of major investments in memory.”

Generally, China’s domestic foundries offer trailing-edge processes on several fronts, such as logic, mixed-signal and RF. In fact, China’s foundries are carving out a decent niche for these processes. For example, Shanghai Huali is currently sold out of 40nm capacity. “There is a lot of demand in the market,” said Zhibiao Mao, a member of the technical staff at Shanghai Huali.

Long term, though, China’s foundry vendors will need to advance its technology. “(China’s) technology is behind,” Gartner’s Wang said. “The key to China’s success is that they need to develop more advanced processes. That way, they can win market share on more advanced nodes rather than using legacy nodes to win share.”

What frustrates the Chinese government is that for leading-edge chips, Chinese foundry customers must rely on the multinational foundries, not the domestic vendors. For example, Semiconductor Manufacturing International Corp. (SMIC), China’s largest foundry vendor, is currently shipping the most advanced process among the domestic firms—a 28nm technology. In comparison, some multinational foundries are ramping up 10nm and/or 7nm finFETs.

Hoping to advance its logic technology, SMIC, Huawei, Imec and Qualcomm in 2015 formed a joint R&D venture in China. The venture plans to develop 14nm finFETs by 2020. “In 14nm, we will try to bring up production earlier that,” said Tzu-Yin Chiu, chief executive of SMIC, in a conference call.

SMIC, meanwhile, is seeing strong demand for mature processes like 65nm, 55nm and 40nm. The demand is fueled by several chip applications, such as digital TVs, set-top boxes and RF.

To meet future demand for these and other technologies, SMIC is building at least three new fabs. In October, the company announced it will build the world’s largest 200mm fab in Tianjin, a port city in northeastern China. At the time, SMIC also broke ground on a new 300mm fab in Shanghai. And in November, it launched a 300mm line in Shenzhen, a city in southern China. It’s unclear what processes will run in these fabs.

Another vendor, Shanghai Huali, recently broke ground on a site that could house three new 300mm fabs in Shanghai. Initially, it plans to build one fab, which will produce 28nm processes. The fab, to cost $5.9 billion, will move into pilot production in the second half of 2018.

Several years ago, meanwhile, multinational foundries entered the China market. For years, TSMC has operated a 200mm fab in Shanghai, while UMC runs a 200mm plant in Suzhou, a city west of Shanghai.

Recently, the multinationals began building new foundry capacity in China with some help. Several Chinese municipal governments have provided a number of attractive incentives to the multinationals. The goal is to bring more chip manufacturing into China. And customers from China and elsewhere want to source more chips within China.

Late last year, UMC moved into production within a 300mm fab venture in Xiamen, a port city on China’s southeast coast. The partners in the fab venture, dubbed United Semi, are China’s Xiamen Municipal People’s Government and Fujian Electronics & Information Group.

Initially, United Semi provides 55nm/40nm processes in the fab. The fab is being expanded from 3,000 to 11,000 wspm by year’s end. It will produce 28nm in the future. “This site will ideally position UMC to capitalize on the vast business opportunities within China’s semiconductor market, while bringing us closer to our Chinese customers,” said Po Wen Yen, chief executive of UMC, in a recent conference call.

Officials from UMC added: “We follow the global capacity situation very closely, including in China. We believe there will be tremendous market opportunities in China from consumer wireless and IoT applications, which require technologies that exhibit a high return on cost versus performance. 40nm and 28nm are ideal candidates for these market sectors; hence, we are expanding our 40nm capacity at Fab 12X and will start 28nm when we receive Taiwan government approval since UMC has proven its high volume production capabilities for these nodes.”

Then, last year, TSMC formed a partnership with the municipal government of Nanjing, the capital of China’s eastern Jiangsu province. TSMC will build a 300mm fab in Nanjing at a total investment valued at $3 billion. The fab, to produce 16nm finFETs, will move into production in the second half of 2018.

The fab is being constructed in two phases. In China, there is a demand for leading-edge technology, including from China’s fabless design houses, according to Mark Liu, president and co-CEO of TSMC. “There is a need for our technology,” Liu said.

Meanwhile, GlobalFoundries last year announced a memorandum of understanding to form a joint fab venture with the government of Chongqing, a provincial-level municipality in southwestern China. It appears that effort was scrapped, however.

Instead, GlobalFoundries and the Chengdu municipality recently announced plans to build a 300mm fab in Chengdu, the capital of the Sichuan province in southwestern China. The total investment for the fab is about $10 billion, analysts said.

Initially, GlobalFoundries will transfer its 180nm/130nm technology from its Singapore fab to Chengdu. It will produce those processes in Chengdu by 2018. Then, in 2019, the fab will produce its 22nm FD-SOI technology. GlobalFoundries also is expanding its fabs in other regions. “These new investments will allow us to expand our existing fabs while growing our presence in China through a partnership in Chengdu,” said Sanjay Jha, chief executive of GlobalFoundries.

“China is the largest and fastest-growing market for semiconductors in the world. Global manufacturing scale is a critical pillar of (GlobalFoundries’) strategy. China is the next logical frontier. It is a huge market opportunity and we need to participate. Chinese customers are increasingly preferring to source chips from China and we are being responsive to market requirements,” added officials from GlobalFoundries. “We are seeing strong global demand for our differentiated technologies, especially 22FDX. Our customers need more capacity and an internal second source to supplement our current capacity in Dresden.”

And if that isn’t enough, Powerchip, a Taiwan specialty foundry vendor, recently formed a fab venture in Hefei, the capital of the Anhui province in eastern China. Still others, including TowerJazz, are also exploring the China market.

Time will tell if China will build all of the proposed fabs. In the meantime, though, it will remain a hotbed of activity, if not the center of attention for the foreseeable future.

Related Stories
China Unveils Memory Plans
Government and industry are investing tens of billions of dollars, but so far results are mixed.
What China Is Planning
With about $120 billion in funding at its disposal, China is looking to buy key technologies it doesn’t already have.
Fab Investment Increases In China
Construction is up, fueled by IoT and internal markets, but the outcome is far from clear.
Sorting Out Next-Gen Memory
A long list of new memory types is hitting the market, but which ones will be successful isn’t clear yet.
The Future Of Memory (Part 3)
Security, process variation, shortage of other IP at advanced nodes, and too many foundry processes.