Tech

Tim Culpan is a technology columnist for Bloomberg Gadfly. He previously covered technology for Bloomberg News.

TSMC Chairman Morris Chang is many things. He's an MIT alumnus, and a Stanford PhD graduate. He's a man known for choosing his words, to which the semiconductor industry listens. As the founder of the world's biggest outsource chip manufacturer, Chang has instilled in the Taiwanese company a U.S.-like paranoia for intellectual property protection and technological secrecy.

What Chang is not, is brash or foolhardy.

So when Taiwan Semiconductor Manufacturing Co. this week signed an investment agreement with a municipal government in China -- a country notorious for IP leakage and lax enforcement of trade secrets -- you can be sure Chang didn't make the decision lightly.

At its site in Nanjing, which it expects to open by the end of 2018, TSMC will make chips using one of its most advanced production technologies. That technology, at a manufacturing scale of 16 nanometers (smaller is better), is the same as the Taiwanese company uses today in making Apple's chips for iPhones and iPads.

Admittedly, the technology will no longer be cutting edge by the time it's in use at TSMC's shiny new $3 billion factory.  So why then did TSMC decide to set up shop in a country where rivals have tried, and succeeded, in stealing its secrets?  

``We aim to provide closer support to customers as well as expand our business opportunities in China in step with the rapid growth of the Chinese semiconductor market over the last several years," TSMC said in Monday's statement.

That China's chip industry is growing is a truism. Consumption of semiconductors in the world's largest PC and smartphone market continues, while the country's ability to manufacture those chips has been unable to keep pace:

Chip Spread
China's domestic semiconductor production can't keep up with demand
Source: PwC, CCID, CISA
Note: 2015-2017 are forecasts.

So, on the surface, that line about providing ``closer support to customers'' seems to make sense. Well, not really. 

TSMC gets 69 percent of its revenue from North America via the industry's biggest names, including Apple, Qualcomm, Broadcom, Nvidia and even Intel. Yet, its presence in the U.S. is limited to one 20-year-old factory while most of the cool stuff is being done in Taiwan. 

With TSMC controlling more than half of the market, not being close to customers doesn't seem to have been a disadvantage. Indeed, with Chinese clients' contribution to sales barely budging, it's hard to make a case that there is a pressing need to be close to them.

No Help from China
Chinese clients have contributed little to TSMC's actual sales growth in the past two years
Source: TSMC data, Bloomberg calculations

TSMC is no Foxconn, either, so moving to China won't help costs. In fact, TSMC CFO Lora Ho admitted in January that ``the cost in China will be still higher than Taiwan'' while scale will be smaller.

The reality for TSMC is that, all things being equal, spending $3 billion on a factory in China doesn't make economic sense. Only a small portion of its Chinese clients -- who design chips to rival Qualcomm et al. -- will even have the capabilities to take advantage of TSMC's 16-nanometer technology when it does become available.

But China doesn't play equal, and TSMC has recognized this. Before it would consider a move to China, it first needed to convince the Taiwanese government to relax an arcane rule preventing local chip makers from setting up new, wholly owned factories across the Taiwan Strait. Under the old law, TSMC would have been forced to jump into bed with a local partner, something that the Taiwanese government soon realized wasn't a great idea.

What TSMC also recognized is that building a local chip industry is of critical importance to the Chinese government, so much so that it's willing to cut outsiders from the market.

``We say that with some degree of assurance from the authorities, some degree of assurance that building a plant there will indeed enhance our access to the Chinese market," Chang told investors in January. ``And reversely, not building a plant there will not enhance."

At the same investor conference, the 84-year-old also backed away from an earlier pledge to grow revenue 10 percent annually for the next few years. Not getting access to China, where growth remains relatively brisk, is a risk he can't take.

It's not only TSMC that needs access. Qualcomm is among clients facing its own troubles in China and getting its most-advanced chips made in-country would go a long way toward placating authorities there.

Chang is famous for anticipating, and meeting, the needs of his global customers, who can't risk being excluded from China either. It's not the carrot of a huge Chinese market that's prompting TSMC's $3 billion bet, but the stick of being frozen out.

This column does not necessarily reflect the opinion of Bloomberg LP and its owners.

  1. By way of comparison, SMIC, a TSMC rival and arguably China's most-advanced contract chipmaker, has yet to get even 1 percent of its revenue from a technology TSMC first released almost five years ago. 

  2. In one of the industry's great ironies, SMIC was caught stealing TSMC secrets and ordered to pay reparations. Because SMIC lacked the cash to pay up, TSMC in 2009 became an unlikely (and unwilling) SMIC shareholder.

  3. If TSMC wants to fill that factory, it may even need to get U.S. clients to agree to have their chips made in Nanjing, despite the risk of IP leakage.

To contact the author of this story:
Tim Culpan in Taipei at tculpan1@bloomberg.net

To contact the editor responsible for this story:
Katrina Nicholas at knicholas2@bloomberg.net