Construction is up, fueled by IoT and internal markets, but the outcome is far from clear.
By Mark LaPedus & Ed Sperling
Fab construction in China is heating up, driven by real and projected demand for IoT devices and the government’s push for internally manufactured chips.
GlobalFoundries, UMC and TSMC are all actively building up fab capacity inside of China, usually in conjunction with other local governments or companies. In addition, SMIC, the largest indigenous foundry, is pushing for its own piece of the manufacturing pie at advanced nodes.
“There’s a lot of interest in building in China, and it’s coming from the government,” said Adrienne Downey, director of technology research at Semico Research. “They’re doing a good job of bringing in outside expertise to make that work. They realize they don’t have expertise in-country, and that this will make them more successful.”
In the 1970s, China had several state-run chipmakers, but those concerns were far behind the West in terms of IC technology. So starting in the 1980s, China began to embark on several initiatives to modernize its IC industry.
The latest effort was launched in 2014, when China unveiled what it called the “National Guideline for Development of the IC Industry.” The plan is to accelerate China’s efforts in a number of areas, such as 14nm finFETs, advanced packaging, MEMS, memory, and various IoT-related processors.
To accomplish these goals, China has created a multi-billion-dollar fund, frequently referred to as “The Big Fund.” The money will be used to invest in its domestic IC firms, as well as to acquire foreign IC companies as a means of gaining quick access to technology.
And as part of that strategy, China continues to lure multinational chipmakers to invest in the nation. Those investments by multinationals are increasing, and for good reason. “China is the fastest growing semiconductor market, with more than half of the world’s semiconductor consumption and a growing ecosystem of fabless companies competing on a global scale,” said Sanjay Jha, chief executive of GlobalFoundries.
Meanwhile, the Chinese government is taking several approaches to lure the multinationals to invest in China. Some strategies are successful, others are not. For example, China’s Tsinghua Unigroup, a state-run venture capital concern, last year launched an unsolicited bid to buy Micron Technology for $23 billion. That deal stalled out amid national security concerns in the United States.
In some cases, though, China’s venture firms and companies have successfully acquired several foreign vendors in the IC sector. The most notable example was STATS ChipPAC, a Singaporean IC-packaging provider that was acquired by China’s Jiangsu Changjiang Electronics Technology (JCET) in 2014.
While China’s acquisition efforts are having mixed results, the nation may have found a blueprint for success. For some time, several Chinese municipal governments have provided a number of attractive incentives in an effort to lure the multinationals into China.
In 2014, for example, Taiwan’s UMC formed a 300mm foundry venture company in Xiamen, a port city on China’s southeast coast. The partners in the fab venture are China’s Xiamen Municipal People’s Government and Fujian Electronics & Information Group, which illustrates how multinational companies are meeting the Chinese government’s mandates for more domestic manufacturing.
Initially, UMC will provide 55nm and 40nm processes in the fab, with plans for 28nm at some undefined future date. UMC anticipates its investment could reach approximately $1.35 billion in the next five years, with a new 300mm fab online by Q4. The foundry has an 8-inch fab outside of Shanghai, as well.
The big drivers, according to Walter Ng, vice president of business management at UMC, are the domestic Chinese market and the expected growth in IoT devices.
“We’re seeing companies in China trying to position themselves to play in the IoT space,” said Ng. “We also believe that we can leverage our location with existing customers in China, including some from the United States. Our new fab isn’t exclusive to the IoT, but it is certainly one of the driving factors.”
A new 300mm fab relieves some of the pressure for 200mm capacity, which is in tight supply, while opening the door for more advanced designs.
“We’re predicting that in the second half capacity will be tight, even with a reduced forecast for the big systems guys,” Ng said. “But it’s better to target for 300mm than 200mm. For one thing, it can handle RF-SOI for switches and other applications. The forecast in demand for RF-SOI is growing exponentially, and there isn’t much 200mm capacity to support growth in switches.”
RF-SOI, the RF version of silicon-on-insulator (SOI) technology, is used to make antenna switches and other components in cell phones. Based on 180nm and 130nm processes, RF-SOI combines CMOS with a highly-resistive, thick-film SOI substrate.
GlobalFoundries, TowerJazz, STMicroelectronics and others also provide RF-SOI processes for foundry customers.
In early 2016, TSMC formed a partnership with the municipal government of Nanjing, the capital of China’s eastern Jiangsu province. Under the plan, TSMC will build a 300mm fab in Nanjing at a total investment valued at $3 billion.
TSMC’s proposed Nanjing fab represents a major change in strategy for the foundry giant. For years, the company has owned and operated a trailing-edge 200mm fab in Shanghai.
In Nanjing, however, TSMC hopes to ramp advanced 16nm finFET processes in the fab by the second half of 2018. This facility will serve a new and emerging crop of leading-edge foundry customers in China, according to Mark Liu, president and co-CEO of TSMC. “The designers are very aggressive in China,” Liu said in a recent interview.
Now, GlobalFoundries is the latest multinational foundry vendor to invest in a fab within China. This week, GlobalFoundries announced a memorandum of understanding to form a joint fab venture with the government of Chongqing, a provincial-level municipality in southwestern China.
Under the terms, the Chongqing government will provide the land, buildings and the infrastructure from an existing 200mm fab in the area. The existing 200mm fab was originally built by Taiwan’s ProMOS in 2011. ProMOS is apparently no longer involved with this fab.
As part of the plan, GlobalFoundries will retrofit the China fab into a 300mm facility, which is slated to move into production in 2017. The company will provide the tools to upgrade the fab to 300mm. The entire site is 400,000m2 with a production cleanroom of 24,000m2.
Initially, GlobalFoundries will use the fab to supplement its capacity from its Singapore facility, which generally develops specialty processes. The company will transfer its 180nm and 130nm processes from Singapore to the China fab.
With the technology, GlobalFoundries will focus on the development of analog, power and mixed-signal processes. Applications include battery management, audio amplifiers, microcontrollers, AC-DC converters and LED lighting.
GlobalFoundries, which did not disclose the fab investment in China, has not ruled out the possibility of developing advanced processes in the fab. “This JV is the beginning of what we expect to be a long-term relationship with Chongqing,” according to officials from the company. “That may include the possibility of manufacturing on more advanced nodes in the future.”
In addition, GlobalFoundries is investing in expanding its design support capabilities in China. At present, the company has design centers in Beijing and Shanghai.
SMIC and other players
Meanwhile, China’s domestic foundries are not standing still. For example, Semiconductor Manufacturing International Corp. (SMIC), China’s largest foundry vendor, hopes to close the gap in process technology with its foreign rivals.
Today, SMIC is ramping up 28nm planar technology, but the ultimate goal is to develop 14nm finFET technology by 2020 or sooner. To help its cause, SMIC, Huawei, IMEC and Qualcomm last year formed a joint R&D venture within SMIC’s fab in Shanghai. The venture will develop 14nm finFET technology.
And others continue to update their investments inside of China, viewing it as a lucrative market over the long term:
• Intel is converting its existing chipset fab to NAND and 3D X-Point memory.
• Silex Microsystems is building a 200mm MEMS fab in Beijing. The company, which was based in Sweden, was bought by Hong Kong-based holding company GAE last July.
• Taiwan’s Powerchip Technology has established a joint venture with the city of Hefei for a 300mm fab to build LCD drivers.
Hiccups in the foundry business
While the highest margins in the foundry business are at the leading edge of design, providing the yield is good, the real money is at older nodes using fully depreciated fabs. This makes the Chinese market a potential boon for foundries, particularly as the IoT begins ramping because most of those chips will be developed at 28nm or above.
But there are some unusual changes that are raising new uncertainties across the foundry business. One involves consolidation at the leading edge of design. Forecasts prior to the combination of companies such as Microchip/Atmel and Avago/Broadcom, are far different from post-acquisition. Industry insiders say a fair amount of business has disappeared, including some businesses that were healthy prior to acquisition.
“It’s a lot more difficult to anticipate what will happen,” said one foundry insider, who asked not to be named. “In the past, if you were using an 8-inch fab the migration path was to 12-inch, and new applications would always come along to take up that 8-inch capacity. It may become more challenging to fill that 8-inch capacity in the future. Even automotive and other IoT applications are focusing on 300mm.”
That creates market shifts and opens the doors to change, and that hasn’t been wasted on Chinese foundries. The question now is just how big of a force China will become in manufacturing, and that may depend on a number of factors.
“China has the potential to become the biggest manufacturing center, but it’s not going to happen overnight,” said Downey. “It is not easy to fill an advanced technology fab. They seem to be leaning toward memory again, in the case of Intel and XMC. That is one way to get into more advanced markets. I don’t think they can ‘take over’ in the next five years. Even if they do, it will take longer than that.”