Wednesday, April 13, 2016

Focus on Japan - How The Semiconductor Industry Collapsed

I mention Japan here fairly often, virtually always in the context of the economic mess the country is in; just last Friday, for instance.  If you were over 10 in 1980, you probably remember when everyone thought that Japan was going to take over the world.  When Sony bought Columbia Pictures and the Rockefeller Center in New York City, it was just taken as more evidence that they were going to roll over the US.  The economic juggernaut was extremely strong, and American businesses adopted such fads as Japanese Management, Quality Circles, and (more recently), Lean Manufacturing.

One of the beliefs at the time was how Japanese electronic manufacturing companies like Panasonic and Sony were going to bury the rest of the industry, so it's a big indicator of just how bad it has gotten in Japan that the semiconductor industry has shrunk to a tiny portion of what it was.  EE Times goes into the story in “Sayonara Japan Semiconductor, Inc”.  They sum it up nicely in this one graphic showing the top 10 semiconductor manufacturing companies in the world by sales.  The ones with a blue background are Japanese companies.  In 1990, six of the ten (!) were Japanese, gathering 38% of worldwide semiconductor sales.  Last year, only one Japanese company remained and it was in 9th place, accounting for only 2.7% of worldwide semiconductor sales.    
The article, while a bit heavy in industry acronyms, is more a business article than a techy piece, and the focus is on a postmortem “whut happened?”.  The generally crappy economic picture in Japan enters into the discussion, but fairly far down the list.  They attribute most of the problems to the celebrated Japanese Management being somewhat less than “genius bar” level.

First, Japanese companies tend to have a cultural obsession with producing excellent products, believing the excellent product will always win the marketplace.  It's hard to argue that's a bad position!  The problem is that the consumer electronics business is incredibly fast-moving and missing a market window while the company tried to make a good product into an excellent one is a mistake that can't be made often.  The company may only get one mistake.  Currently, for example, Sony's CMOS image sensors are used in many digital cameras but aren't necessarily cost effective.  If Sony isn’t doesn't watch that vigilantly, Omnivision, acquired by China’s Hua Capital Management earlier this year, and Galaxycore, an up-and-coming Chinese CMOS image sensor vendor, could mean pain coming for Sony.

Similarly, Toshiba and their Flash memory.  Right now, Toshiba is a major player in the Flash memory market.  Change is coming to that world: we mentioned the trend for Flash memories to start going vertical here in January, and Toshiba needs to put in the massive capital equipment investments needed to go vertical if they're going to keep up.  Read that as a question of whether management can accomplish the investment.
Akira Minamikawa, director, semiconductor value chain at IHS Technology, remains similarly cautious of Toshiba’s NAND future. The IHS analyst sees “at least a one-year time lag” in Toshiba’s 3D NAND flash memory, compared to Samsung. “Sure, Toshiba says its 3D NAND is sampling now. But there are usually many more steps between sampling and volume production.”
But without a doubt, the consultants interviewed for the piece all said the main culprit was management.
IC Insight’s Matas said, “Japanese business leadership that takes the long-term view couldn’t deal successfully with rapidly changing global market needs and dynamics. The ‘hot’ item in today's consumer market is tomorrow's old news. It’s partly culture-based, but Japanese semiconductor companies have been too slow to adjust.”

IHS analyst Minamikawa said, “The top executives at Japanese firms stuck to the idea that as long as their companies have superior technologies, they wouldn’t lose.” They turned a blind eye to factors — other than technologies — that would make the chip business successful.
A big sign of the problems with management was R&D investment that never resulted in marketable products
Yunogami observed that every national project and consortium built around the development of semiconductors in Japan over the last two decades has failed. Instead of helping, they made Japanese chip companies “gravely ill,” he said. “In some cases they literally killed a few Japanese chip vendors.”

Japan has never stopped investing in R&D. “Japan loves R&D,” he said. But, as things have turned out, every prescription developed by the Japanese government was either dead wrong, or out of step with the global market reality. Making matters worse, Japan’s top managers failed to lay out the steps it must take to restore their companies’ profitability, Yunogami added.
The failure of management includes the failure to work effectively with major companies, in particular System on a Chip (SoC) buyers.
There is a fundamental structural issue in the Japanese semiconductor industry. Chip vendors in Japan took great pride in designing SoCs, but those designers have been often too dependent on customers (or the system division of their parent companies) to design a system, and develop architecture for SoCs, Yunogami explained. In fact, [Chipmaker] Renesas relies on Toyota and Denso to develop much of the value-added portions of its SoC designs. Renesas is just their chip supplier.  [Note: clarification added - SiG]

Too many Japanese SoC designers are accustomed to twiddling their thumbs until their customers bring in designs, Yunogami concluded. “Japan lacks initiative.”
Japan's strength was once thought to be their "vertical integration"; how producing both the low level parts for their own top level box (be it a Sony Walkman or a Panasonic Toughbook) gave them advantages over American (and other) designers buying generic parts and trying to develop innovative parts with ordinary parts.  No one says this anymore. 
“Japan's once strong, vertically integrated business model fell apart and  missed out on smartphone growth — perhaps the greatest opportunity since the PC market.”
The relentless push to put smaller and smaller features onto a silicon wafer to put more and more parts in the same area makes capital investment imperative for the semiconductor manufacturers.  These are extremely expensive machines.  For perspective, the current 20 nanometer scale parts equates to about 180 silicon atoms in a row and the photolithography equipment they use needs to hold that faithfully.  The step to 10 nm parts will occur in "months, not years".

Without a doubt, the consumer electronics business has to be among the most difficult to manage.  In a quick turn industry, the first chip maker to develop a way to accomplish something will charge high prices.  The next company undercuts their price, so the first chip maker must respond.  As additional players come into the market, prices fall to a floor all the chip makers can meet, and the end customer's product gets very price sensitive.  Getting into the top 10 companies in the world is rough.  Getting knocked out of the top 10 can be a single mistake, or a few minutes of inattention. 


  1. Very good article.

    You know, if Japanese management principles turned out so badly for Japan, why are so many US companies embracing Lean Manufacturing? Is it that the US companies have actually made the principles work because of the cultural differences? Americans (well, some) have initiative, creativity, and resistance to stagnation that (some) Japanese lack, because of the differences between American culture (well, part of it) and Japanese culture (the older generation). Thoughts?

    1. I think of Lean as a way of optimizing what the company does, and what this article was implying was that they didn't need to do that, they needed all new methods. Once they developed the new technologies, then they could repeat the lean optimizations. They needed to go past what lean could do for them.

  2. It sounds to me like the methods were not enough in fast changing industries - particularly in some of the popular internet ares (think apps, cell phones, social media) hype and marketing 'splash' matter more than quality.
    When applied to a slower industry, they can help with lower costs and consistent quality, right?

  3. To some extent Japan's rise to economic power was based on their ability and will to copy someone else's work. In the 60's some of the Japanese motorcycles were virtually identical to English and American bikes such that engine components were interchangeable. Many of Japans big heavy equipment units are indentical to Caterpillars machines. This 'skill' lends itself to the slower more stable industries.

  4. One of my college instructors told me something very interesting one day.
    He said flash memory was very expensive because it was made to such exacting standards.
    The size of flash memory chips stayed small because of the price point.
    Then he said someone came up with a breakthrough that revolutionized flash.
    They noticed that they had a bunch of rejects that could be made useful by writing a program that would map the bad sectors and only use the good ones.
    Then flash became cheaper and people wanted larger and larger flash memory chips.
    And flash supplanted other data storage methods as a result.
    What I wonder at is how long it took for someone to come up with the mapping idea.
    Electrical engineering can be very precise.
    I marvel for instance at the precision of components like high speed oscillators that can be precisely tuned to very fine radio frequencies.
    And yet, it is based on the adaptation of ideal world models to the real world.
    We accept that ideal diodes don't exist because all diodes have a reverse current.
    All capacitors leak, some more than others.
    All inductors are resistive.
    Everything is imperfect in electrical engineering.
    So, why was flash memory manufactured to perfection for so long when we work at managing the ideal versus real every day?
    In the context of your post I see a parallel between innovation in flash memory and the decline of the Japanese semiconductor industry.
    Yes, quality is good.
    Perfection is aspirational.
    And ultimately artisanal.
    I can buy a crafted kitchen knife that will outlive me for a few hundred dollars.
    Or I can buy a cheaper one that I might have to sharpen a few times and maybe replace every once in a while.
    On the other side of the electrical engineering Laffer curve are the Chinese.
    Poor quality plunbing the depth of what people will accept.
    I am sure, however that both the Japanese and the Chinese make perfectly imperfect write only memory.

    1. My Standard Disclaimer here is that I'm not a digital hardware designer, but I've been told it's actually even a bit more involved than that. Flash memories go bad after some number write cycles, so that sector mapping is needed for operational use, too. A digital designer I just worked with told me that Flash chips are sold to a smaller capacity than they really have. The sector mapping utility is kind of an Operating System on a chip for the Flash. It keeps track of what areas it can and can't write to.

      When you buy a solid state drive, they keep track of bad sectors and at some point, the drive is considered dead. That point is lots of write cycles away.