Japan’s DRAM Makers
By David Lammers
The announcement that Elpida Memory Inc. is declaring bankruptcy essentially takes Japan out of the DRAM sector which Japanese companies came to dominate in the mid-1980s.
What happened in Japan that took out all the once-leading DRAM makers over three decades?
Three things come to mind: First, the yen tripled in value vis a vis the dollar. Secondly, Japanese managers lacked strategic direction and became overly cautious. Third, Samsung and others, sensing opportunity, pounced on the opening.
Go back 30 years to the lovely, green island of Kyushu in southern Japan. Toshiba embarked on a gamble there — one that eventually earned the company billions of dollars in profits — moving from NMOS to the more power-efficient CMOS transistor technology. At Toshiba’s Oita Works in the early 1980s, Toshiba DRAM engineers had a single, parts-starved implanter and were staying up all night, starting at 11 pm every night, to boost the single-digit yields of their one megabit CMOS DRAM.
By the mid-1980s, Tsuyoshi Kawanishi, the general manager of Toshiba’s semiconductor operations, and his staff had taken yields for the 1-Mbit CMOS DRAM to respectable levels. Kawanishi was a charismatic manager, confident without being arrogant, who knew that CMOS at the megabit density was a winning combo in a memory IC sector that was growing rapidly as the computer boom got fully underway. His staff idolized him and – similar to the positive momentum built by Jim Morgan at the helm of Applied Materials – the positive wave at Toshiba carried over into successes in notebooks computers, displays, and other products.
An equally talented Toshiba semiconductor division manager, Taizo Nishimuro, was at that time in charge of sales and marketing. He was making sure that Toshiba’s limited supplies of CMOS DRAMs were allocated fairly to the world’s computer makers, which were all were clamoring for parts. “We know where every single one mega DRAM is going. We have to, because of the trade frictions,” Nishimuro told me in an a 1986 interview.
Nishimuro-san was right to be worried. In 1985 the United States Dept. of Commerce had slammed DRAM dumping duties on Japan, setting off a series of events which led to the 1986 U.S.-Japan Semiconductor trade agreement, which formally called for an expectation that U.S. companies would hold a 20 percent share of an open Japan chip market.
It took a year for NEC, Hitachi, Mitsubishi Electric, Fujitsu, Matsushita Electric, Oki Electric, and others to catch up with Toshiba’s lead in CMOS DRAMs. But they did it, and Japanese semiconductor companies came to hold six of the top 10 spots in the world semiconductor rankings. At one point, investments in Japan’s semiconductor fabs were more than half of all Japan’s capital investments – an astounding amount of money was going into memory fab building. It was enough to scare Intel and other U.S.-based DRAM makers out of the memory sector and into MPUs – a fortuitous change in direction for Intel.
As exports of DRAMs and other well-made products soared, Japan’s trade surpluses ballooned. The imbalances were made worse by the country’s shima kuni konjo (literally, island country mentality, which came to mean an arguably insular belief by many Japanese that nearly everything done in Japan, by Japanese people, was more reliable and more valuable than products made outside of the Japanese islands). With an anti-import bias entrenched in the national mentality, Japan’s trade imbalance proved nearly intractable.
As a result, the dollar, worth about 240 yen in early 1985, quickly started rising in value to the 120- 130 range. The yen now regularly trades at 75-80 yen to the dollar, which is in sharp contrast to the relatively stable rate seen for the Korean won over the past decade.
Thus, just a few years after Toshiba’s success with its CMOS megabit DRAM, Japanese managers had twin headaches: how to deal with the higher yen and the insistent market-opening demands from the U.S. Semiconductor Industry Association, led by Bob Noyce, Gordon Moore, Wilf Corrigan, Jerry Sanders, Charlie Sporck, and other battle-hardened executives.
The persistent trade imbalances which Washington and the SIA railed against led to en daka (too high value of the yen). The yen’s value became a ever-heavier weight bearing down on the Japanese semiconductor industry.
As the yen was rising, the confidence of Japan’s top managers – who had computer and telecom operations as well as semiconductors to deal with– began sagging. The executives leading Japan’s electronics industry moved away from a stance of bold initiative — exemplified by Toshiba’s Kawanishi and the CMOS move – to a more hesitant, bottom-line-driven mentality.
Again, cultural perceptions played a role. During this early 1990s period I interviewed Kazuo Kimbara, who ran Hitachi’s semiconductor operations. He fretted that the United States economy would be dragged down by social challenges such as crime, education, racial conflicts, divorce, and the like. Kimbara-san, and others, couldn’t quite see beyond the headlines to the basic realities of the U.S. economy, which proceeded to do quite well as software, networking, personal computers, and mobile products combined to form the tech-driven economy we have today.
As Japan’s confidence and chip investments faltered, Samsung pounced. Everything broke right for Samsung, as the Korean won remained relatively stable while the yen rose sharply in value. Investments in and revenues from NAND complemented its DRAM base. And Samsung invested during the industry down cycles, and in 200-mm and then 300mm wafer equipment while Japanese companies did not, relatively speaking.
Another point has to do with the sequence of development of Japan’s semiconductor industry. Hajime Sasaki, who ran NEC’s semiconductor operations in its heyday, famously said Japan’s chip industry was “like a dog chasing two rabbits” — logic and memory — ending up with neither. Samsung seemed to uncoil like a baseball slugger, initially seizing the DRAM opportunity, then building into NAND, and – having built a solid base in commodity memory – moving on to the logic and foundry opportunity.
A few words about Elpida. Its president, Yukio Sakamoto, was a vice president at Texas Instruments Japan until 1993. TI operated several fabs in Japan, and Sakamoto moved up the ladder by solving tough problems, recalled Mentor Graphics CEO Wally Rhines, who also worked at TI then.
“When a customer would call up and say it needed a rush shipment by Friday, Sakamoto would tell us, ‘I’ll take care of it.’ And he would. He did that over and over again, saying ‘I’ll take care of it.’ That is how he gained everyone’s respect, by taking care of the tough challenges that come up when you are running a semiconductor operation,” Rhines said.
As talented and committed as Sakamoto-san surely is, he and his colleagues at Elpida ran up against a set of new realities in the DRAM industry. They were among the leaders in the drive to 20nm design rules, to Wide I/O DRAM and Mobile DRAM. Elpida remains a major supplier to Apple. But is it enough? Will the Japanese government move to save Elpida and its jobs, or let it go just as Germany let Qimonda die?
Without NAND and facing a bleak future for DRAMs, Elpida may in fact go bankrupt. Japanese chip manufacturers simply can’t thrive in an era of 78 yen to the dollar.
Barring a phoenix-like miracle by Elpida, it is a sad end to a remarkable run by a long line of Japanese DRAM makers.