SEMICON China: Is This China's Decade?

One of the presentations on the opening day of SEMICON China, here in Shanghai, was by Rick Wallace, who is the CEO of KLA-Tencor (KT). He started by pointing out that KT have been selling for the entire history of the industry, so he thought it would be good to give a perspective on the growth of the industry and how it has moved around geographically over the decades.

He titled his presentation The Geographic Evolution of the Semiconductor Industry: From 1970 to 2020.

He started with a graph summarizing the growth of the semiconductor industry over the last five decades. On the left is the industry revenue. On the right is KT revenue by location, and you can see the support of the ramp in the different regions. For KT today, most of their business is in Asia.

The industry started in Silicon Valley in the 1970s. Despite the name, there's not a lot of silicon in Silicon Valley anymore. The last fab closed about ten years ago, the Intel TD line. My experience at VLSI Technology was typical. We had our first fab on McKay Drive in San Jose. However, we never had another fab in Silicon Valley. The second fab was in San Antonio, Texas. Various joint venture fabs were planned in East Germany and Malaysia, but somehow they never got off the ground before VLSI was eventuallly acquired by Philips Semiconductors (now NXP, of course).

In 1975, in the middle of the 1970s, you can se the top ten semiconductor companies. Two points about these companies. Several are names that still exist, like GE, but no longer do semiconductor manufacturing. Several have been acquired and still exist in some form inside other companies (such as National, which is now part of TI). Only one company, Intel, still does leading-edge manufacturing. This was an era when, as Rick put it, "You could drive across the valley from KT to meet your customers". That's a bit of an exaggeration, Dallas is a long drive, but the point is that in the 1970s, almost the entire semiconductor industry was domestic.

In the 1980s, it was the era of the PC, which meant microprocessors and memory. This was the age of Japan when they came to dominate the memory market. They used statistical process control and continuous improvement to improve quality to an unprecendented level. The god in Japan in that era was Edwards Deming, an American, "a prophet without honor in his own home". As an aside, before he died, but when he was in his 80s, I went on one of his courses on quality. Somewhere I still have a signed copy of his famous book Out of the Crisis.

Japan seemed so invincible that, in this era, a consortium of semiconductor manufacturers along with the US government created SEMATCH to address how to survive in the face of Japanese competition. But the Japanese were not invincible, and, of the nine Japanese companies listed, only one still does volume manufacturing. The Japanese semiconductor equipment industry also came into being, some of which still exists today.

In the 1990s, it was the turn of Korea. They decided, like Japan before, that memory would be the route into a competitive semiconductor industry, since it required just a small number of designs and then feeding experience back to drive fast yield ramps. They soon had a very strong position in the market. Samsung, in particular, started the decade as #12 in the semiconductor industry and ended it as #4 (and has since gone on to take the crown).

Or, as Rick put it, they "focused on memory and working hard." The motivation seemed to be first, Korea; second, their company; and their personal success only third.

In the 2000s, Taiwan (or China Taiwan as you have to diplomatically phrase it when you are in China) came up to the top. They also focused on process control (naturally using lots of KT equipment, but Rick left us to draw that conclusion ourselves). Taiwan was a more freewheeling environment than either Japan or Korea, places where people stayed at their first company for life. There was lots of movement, companies had lots of difficulty keeping people, but they created an ecosystem. It was also attractive enough to bring back expats who had gone to the US without any expectation that they would ever return.

TSMC developed a new model that has changed the world (this is a good moment for me to put a plug in for Dan Nenni and my book Fabless: The Transforamation of the Semiconductor Industry, available on Amazon for $20 if you insist on dead trees, and around the web for free if you don't). Until this era, most companies had been successful with just a few devices shipping in very high volume and a process dialed in for success on just those few designs. The foundry model required different equipment to cope with all the different designs, and a diffeent type of process control when it couldn't be tuned for each design.

In one sense, the 2010s are already looking like the decade of China. Certainly, there is an enormous investment, over $100B depending on just what is counted. China has tried to be a force in semiconductor before, but failed when the investment was too spread out. Until recently, people expected fewer and fewer, bigger and bigger fabs. But China is changing that with lots of fabs of different sizes, technologies, wafer diameters, and technologies.

You can see the baton being passed, at least in terms of investment, in the bars above. China has come to dominate investment in semiconductor in terms of number of projects (the biggest projects are stil elsewhere, so in terms of investment dollars it is more evenly spread). This is shown by the red bars that are dwarfing everything else going forward.

They are also spread across many different regions. Sorry the picture above is a bit washed out and you can't really read the labels (ok, you can't read them at all). But you can see the geographical spread of the semiconductor facilities across the country. Rick pointed out that SEMI's mantra of connect and collaborate is important since nobody has all the resources to work across so many projects. Multiple projects, multiple places and multiple products, which makes for multiple challenges. Perhaps the biggest is that hiring people is a challenge, moving people is a challenge.

There are also more prosaic challenges, at least for KT, of getting stuff in and out of the country, which is its own challenge. But Rick has never been as excited as he is today.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.