Wednesday, October 16, 2019
Brakes Put on China’s Semiconductor Ambition
Washington Keeps China in Check
Brakes Put on China’s Semiconductor Ambition
  • By Kim Eun-jin
  • March 18, 2019, 13:49
Share articles

The Chinese semiconductor industry is having trouble reaching its goals in both the memory and non-memory sectors as the United States is stepping up its efforts to keep it in check. 

China’s scheme to foster its semiconductor industry has been going awry. The Chinese semiconductor industry is having trouble reaching its goals in both the memory sector, which is dominated by Samsung Electronics and SK Hynix of Korea, and the non-memory sector where the United States has begun to protect intellectual property rights.

Yangtze Memory Technology Co. (YMTC), which leads the Chinese NAND industry, released a prototype of a 32-layer (3rd generation) product last October. At the same time, the company released a blueprint to go straight to 128-layer products in 2020, skipping 64-layer and 90-layer products. Samsung developed 32-layer NAND flashes in 2014. This means that there is at least a four-year technological gap between Samsung and YMTC. SK Hynix began mass production of 96-layer products in 2018. Samsung plans to launch a 100-layer NAND flash in the second half of this year. If YMTC produces a 128-layer NAND chip next year, its technological gap with Samsung will significantly narrow to about one year.

But there has been no news about technological progress by the Chinese chipmaker. A 3D NAND flash is a set of vertically stacked memories of a planar structure so the higher the number of layers is, the larger its capacity is. 3D NAND flashes are divided into three generations based on the number of layers – the third generation (32 layers), the fourth generation (64/72 layers), and the fifth generation (92/96 layers). It usually takes one year to move to the next generation. "YMTC must be anxious to develop 6th-generation products faster as their newly developed 32-layer products have no cost competitiveness," a semiconductor industry official said. "But we know that YMTC has not even succeeded in mass production of 32-layer products yet. In the face of a downturn of the memory market, YMTC is facing an uphill battle to overcome a situation where development cost balloons."

There is a big mess in the Chinese DRAM industry. The starting point of DRAM mass production by Chinese chipmakers has been delayed since last year, and Fujian Jinhua, which is leading the server DRAM sector, is rumored to be withdrawing from the DRAM business. DRAMs affect system performance. DRAM technology is more sophisticated than NAND technology so the United States is doing everything in its power to protect its DRAM market. This explains why Fujian Jinhua suffered a double whammy –- a U.S. embargo on semiconductor equipment exports to China and Taiwanese foundry maker UMC’s severance of ties with Fujian Jinhua. Even though Fujian Jinhua came up with a prototype of 32-nano DRAMs, which Samsung began to mass-produce in 2011, it is not clear whether the company will be able to mass-produce them by the end of 2019. Experts are also skeptical about mass-production of 25 nm mobile DRAMs by Innotron, a mobile DRAM maker that has a reputation as the Chinese memory maker with the best technology.

As China’s scheme to foster its semiconductor industry has not panned out, it has become inevitable for the Chinese semiconductor industry to focus on the non-memory sector. According to statistics, Korea had 150 fabless companies as of 2016 while China had 1,300. SK Hynix has decided to build a foundry production plant in Wuxi in the second half of this year because the plant will be able to land many orders from fabless companies in China.

But there is also a catch in the non-memory sector. A case in point is the fact that Intel recently said a good-bye to Tsinghua Unigroup of China, which announced its plan to commercialize 5G telecommunication chips within this year, due to pressure from the U.S. government. Experts say that Tsinghua Unigroup faces many difficulties in developing new chips, still less mass-producing them. "This case shows that the United States is expanding its check of China even in the non-memory sector," said an executive of a semiconductor company.

SMIC, a Chinese foundry company, is also having an uphill battle. It is highly unlikely that the company will meet its desired yield in the 14-nano PinPet process until the end of the first half of 2019. Although SMIC can boost its technological power at one fell swoop if it acquires GlobalFoundries, the United States will not sit on the fence, many experts say.