Globalfoundries' Chengdu fab accounts for 10% of China's race for semicon production, but officials believe there is a wider blitz of investment.
In last week’s announcement, Globalfoundries made it clear that after about a year of producing 180/130nm chips using processes transferred from the foundry’s Singapore operations, the second phase, scheduled to begin in 2019, will be devoted to 22nm fully depleted silicon-on-insulator (FD-SOI) chips with technology from Dresden.
Globalfoundries’ announcement last week that it intends to build a $10 billion fab in Chengdu, China, came as no surprise to the industry observers following China’s aggressive pursuit of a local soup-to-nuts semiconductor eco-system.
Bill McClean, president of market-watcher IC Insights, told my colleague Rick Merritt, “With TSMC, UMC and SMIC expanding their foundry capacity in China, I believe that Globalfoundries believes that they need to do something as well.”
Globalfoundries’ move is a reminder of China’s inexorable arms race for semiconductor production capacity.
Of course, every major business/technology announcement brings collateral damage—or advantage—to other companies and/or technologies.
This poses questions. Who just gained from the Globalfoundries’ deal? Who might be losing ground–in the long run–as the fab tsunami sweeps over China?
Let’s start with the winners.
Undoubtedly, the big winners are the promoters of fully depleted silicon-on-insulator (FD-SOI) technology. Their concerted effort to push FD-SOI as an alternative process technology for low-power, mobile devices may be finally coming to fruition in China.
A few telltale signs have consistently linked China with FD-SOI, but nobody was never sure which Chinese foundry, if any, was willing to bet on FD-SOI. The world was waiting for another shoe to drop.
Now, we all know the answer. It’s Globalfoundries’ new fab in Chengdu.
Besides the Chengdu announcement, several data points indicate that China is quietly becoming the development focus for FD-SOI:
Figure 1: Inside Huali's $5.9 billion bet on Fab 2.
When reached by EE Times, Soitec’s spokeswoman said Globalfoundries’ Chengdu announcement “confirms our strategy in the mobility markets.” In Soitec’s view, this means not only that “FD-SOI is becoming a standard just as RF-SOI already is,” but also it “vindicates Soitec’s strategy in China.” Globalfoundries’ investment in Chengdu will be “instrumental to support China’s growing fabless semiconductor industry,” she added.
In short, FD-SOI has officially become a part of China’s semiconductor roadmap.
So, who could possibly lose anything as a result of Globalfoundries’ latest investment?
Government officials in China reportedly believe that Taiwan Semiconductor Manufacturing Co. will feel the pain. TSMC’s loss won’t necessarily be technology, but people, they say.
This conclusion, however, isn’t directly tied to Globalfoundries’ Chengdu investment alone. The Chinese officials believe TSMC will suffer from the broader onslaught of fab investment in China by various parties—including the latest by Globalfoundries.
One of my sources in Silicon Valley said Chinese government officials are telling him to “do the simple math.”
China has committed thus far well over $100 billion for semiconductor manufacturing.
Not included in the table above are additional investments announced in January by Tsinghua Unigroup to build a $30 billion memory chip fab in Nanjing, and Globalfoundries’ $10 billion Chengdu facility.
When we write about new fabs, we tend to think only in terms of what companies must spend to build capacity and procure equipment. But how much will it cost for China, with little experience in actual IC production, to headhunt the necessary engineering talent?
With so many fabs suddenly popping up, China must look for “manufacturing people” versed in “process technology, yield, failure, line management, others,” according to the source. “If headcount related expense is comparable, China would need roughly three times the entire TSMC headcount,” he noted.
This source estimated that TSMC today has less than 1,000 hard-core technology/manufacturing people. Certainly, TSMC possesses unmatched strength in IP. It has a lead–more than five years–over the competition, he estimated. But what if TSMC starts losing its own engineering talent?
Assume that $1 million per head is needed in recruiting—with incentives and bonuses. “It would only cost $1 billion to recruit 100% of all the headcount in TSMC,” the source said. In the bigger scheme of things, this represents only 1% of China’s manufacturing investment budget.
Is it possible for China to merge some of its own local fab operations with TSMC? Could the day come when Taiwan Semiconductor Manufacturing Co. becomes China Semiconductor Manufacturing Co.?
Admittedly, that sounds like a tall tale scribbled on a cocktail napkin at a dinner table where a few guys have sucked down a little too much baijiu. But it’s a scenario with a credible grain of truth.
Figure 2: Fab activities in China.
It turns out the brain drain problem in TSMC is neither pipe dream nor secret. Taiwan’s newspaper China Post reported on Jan. 25, 2017:
The outspoken Taiwan Semiconductor Manufacturing Co. Chairman Morris Chang slammed Premier Lin Chuan on Monday [Jan. 23] for failing to stem a brain drain of technology talent.
Such workers had been poached by China, Chang said during a meeting at the Executive Yuan on Monday evening, who added "the government should think hard on how to keep them and draw more of them to Taiwan."
"Simply offering subsidies is not good enough," Chang reportedly said.
Backing Chang's statement, the China Post story quoted Ko Chen-en, a National Taiwan University business professor, saying that almost half his students had been approached by Chinese companies after graduation.
This is not to suggest that TSMC has turned a blind eye toward China.
TSMC, particularly Chang, has been mindful enough of China to commit to building a fab in Nanjing, just like its competitors.
But this investment will be only $3 billion, much smaller than the competition’s commitments. It’s not clear whether TSMC intended to plunge so shallowly into Nanjing. There are suspicions that TSMC middle management might not have fully embraced Chang’s original intent.
One thing is clear. Although TSMC isn’t losing its technology edge to any of the newbie fabs in China, the Taiwanese foundry giant has a lot to lose. It must find a way to stand up effectively China’s $100 billion spending spree on local chip manufacturing.
First published by EE Times.