SAN JOSE, Calif. — A plan for a billion-dollar fab in Guangzhou, China, aims to pool investments from fabless companies, creating a built-in customer base. It is the latest addition to a growing list of projects as the country tries to build up its semiconductor sector.
The so-called CanSemi project announced plans to build a 300-mm fab capable of producing 30,000 wafers/month. It would be a crown jewel of the Guangdong provincial capital, formerly known as Canton, that has aggressively courted investments from many electronics companies in recent months, including Foxconn and LG.
The CanSemi project, also known Guangzhou Yuexin or YPC, aims to start operations as early as 2019, according to a Dec. 26 report from ESM China, a sister publication of EE Times. Richard Chang, the Taiwan-born founder of China’s largest foundry, Semiconductor Manufacturing International Corp. (SMIC), is said to lead the project.
Chang is also reportedly involved in a related project to create a chip design company in Guangzhou. He initially took his idea for a foundry backed by fabless customers to Ningbo, a city in Zhejiang province, but was unable to get enough support.
CanSemi released few details on the sources of its funding, its technology, or specific product plans. It aims to make chips for the Internet of Things, car networking, artificial intelligence, and 5G, leading U.S. observers to speculate that it could start with a 40- to 28-nm process. So far, it has announced local government and financial companies as investors but not any fabless chip vendors.
China already has 20 chip fabs currently under construction and many more reported or rumored to be in the works, said Christian Gregor Dieseldorff, a research director who tracks global fabs for the SEMI trade group’s research team. Most of the 20 projects that the trade group has verified are China-run, and 70 percent of them are 300-mm fabs, he said.
An artist’s rendering of the proposed CanSemi in Guangzhou. (Image: ESM)
Some of China’s projects face challenges getting funding, access to process technology, or an adequate supply of technical workers, said analysts.
For example, Yangtze Memory Technology Co. (YMTC) was announced with much fanfare, with plans to build up three fabs with a total capacity of 300,000 wafers/month. SEMI believes its current output in a single new fab next to XMC in Wuhan is limited to about 5,000 wafers/month.
Another source who asked not to be named said that the YMTC fab is progressing well and on schedule to make 64-layer 3D NAND flash before the end of the year. Samples of 32-layer parts made at XMC are already said to be available. The company did not respond to a request for information.
Meanwhile, YMTC’s main investor, Tsinghua Unigroup, announced deals to build a total of six more fabs. Those deals aim to create maximum production capabilities of 300,000 wafers/month in both Nanjing and Chengdu. Together with YMTC, they aim to someday make nearly 1 million wafers a month, “an absolutely crazy number,” said the unnamed source.
Unigroup’s first Nanjing fab, announced a year ago, broke ground late last year. Its first Chengdu fab has not started construction, and funding is believed to be incomplete.
Separately, SMIC is still working with Qualcomm and the Imec research institute on its first node supporting 14-nm FinFETs, a process that’s been available from companies such as Samsung and TSMC for several years.
China has never been short of ambitious plans that sometimes don’t come to full fruition. Market watcher IC Insights recently forecast that China will fall far short of government goals for supplying up to 70 percent of the chips that it consumes by 2022.
Next page: A clash for cash to build an industry