Japanese research takes aim at US hold on chip design
Atsugi, Japan — At Fujitsu Corporation's new semiconductor lab a hundred miles southwest of Tokyo, Dr. Takahiko Misugi has just spent 15 minutes ticking off his company's accomplishments in the semiconductor research that forms the backbone of the computer revolution. It is the stuff of countless PhD dissertations: a laundry list of arcane innovations peppered with references to HEMT and CMOST, gallium arsenide and electron mobility rates.
``Japanese aren't supposed to be so good at basic research, right?'' says Dr. Misugi, who punctuates the irony of his question by dropping a small stack of Japanese research papers on the table in front of him. ``Well, we're trying to be better at it than we used to be.'' Taking the long view
Becoming ``better at it'' has almost become an obsession among Japan's semiconductor manufacturers. As the 18-month semiconductor boom tapers off amid widespread concern about oversupply, producers in the United States are pruning their investments in semiconductor plant and equipment by as much as 20 percent.
But not those in Japan. Japanese firms surpassed those of the US in semiconductor capital investment for the first time last year, topping $4 billion in 1984 -- more than either the Japanese car or steel industries, both of which are traditionally more capital-intensive.
The investment gap with the US promises to widen. In the fiscal year that for many companies begins this week, Nomura Securities estimates that market leaders like Mitsubishi Electric, Hitachi, NEC, and Toshiba will be increasing their investment by 15 to 20 percent above the $2.5 billion collectively invested in 1984.
At the same time, most producers are increasing the numbers of researchers studying new semiconductor technologies. A few, such as NEC and Matsushita, have taken the heretofore unconventional approach of importing researchers from overseas.
``We cannot increase our market share by tying ourselves to the short-term fortunes of the market,'' says Akira Harada, executive vice-president of Matsushita. ``We take the long view.''
So far, that approach has been a successful one as Japanese producers have steadily expanded their activities in the semiconductor field. From a position of serious technological disadvantage in the mid-1970s, Japan has arrived at a point where it hosts three of the five largest semiconductor manufacturers in the world: NEC, third behind Texas Instruments and Motorola, is trailed by Hitachi and Toshiba. Japanese companies accounted for nearly 40 percent of the $23.4 billion in worldwide semiconductor shipments in 1984, up from 36.8 percent of the $18.7 billion in '83. Customizing chips
Particularly spectacular have been the achievements of Japanese companies in the area of memory chips. More than 60 percent of these data-storing computer circuits came from Japan, according to Dataquest, a California market-research firm. Yet for all their success in dominating the market for 64k and 256k RAM (random access memory) chips, many company executives say they will have to work harder than ever to maintain their impressive level of performance.
The reason: A trend toward more complicated ``customized'' chip designs for specific computer applications, some observers speculate, could leave many Japanese producers behind. Their strength, the thinking goes, lies with a vaunted production technology -- an aptitude at producing mass quantities of relatively simple chip designs at the lowest cost -- rather than an ability to develop elaborate chip designs for limited production.
``My American friends ask me whether I think the emphasis in chips for the future will be on customization or production,'' says Fujitsu's Misugi. ``I tell them customization, and they always say `then the advantage is with the us.' ''
But industry leaders here are actively seeking to expand their efforts to encompass the more complex chips demanding innovative design. Some efforts are taking the shape of forays into the realm of microprocessor design. Generally, Japanese companies have either bought the plans for US-engineered microprocessors -- basically the ``brain'' of a computer -- or simply ``stolen'' them. Other efforts concern themselves with attempts to design extremely dense ultra-fast circuits for 21st-century computer systems.
But all of them, Japanese manufacturers say, share a common motivation. The companies want to wean themselves from a technology that is beginning to be exhausted as a search for faster and more powerful computer devices continues.
``We are coming to the end stage of the present technology,'' says Hiroyuki Mizuno, semiconductor research director at Matsushita. ``If we fail to develop new technologies ourselves, we will fall behind.'' A new chip generation
The beginning of the final chapters of the present technology -- basically the technique of etching tiny circuits in a bed of silicon -- were written in March when Toshiba announced its readiness to send sample shipments of the much-awaited one-megabit RAM. NEC, Matsushita, Hitachi, Fujitsu, and Mitsubishi all later said that they, too, would have their own one-megabit RAMs ready by the end of the year. Meanwhile, US companies -- AT&T, IBM, and Mostek -- were left in the prototype stage. European companies such as Siemans of West Germany and Phillips of Holland found themselves farther behind.
Having jumped the foreign competition once again, however, Japanese companies say they are not about to rest on laurels. Already they are actively planning for the next generation of memory chips, which promise to strain the capabilities of current silicon-based technologies. ``In this business, it's come to the point where one new chip is nothing more than a stepping stone to the next one,'' says Peter Worlff, an industry analyst with Prudential-Bache in Tokyo.
The next chips would, say those in the industry, be the last in the era of very-large-scale integration (VLSI) -- a reference to the enormous numbers of circuits packed onto a wedge of silicon. They would sport enough memory circuits to store four, or eventually, 16 megabits of data. Beyond these, researchers talk of ultra-large-scale integration (ULSI) where exotic, yet-to-be-developed technologies result in 100 megabit memory chips.
In the pursuit of the ULSI quarry, Japan does not intend to find itself in the position the US and Europe are experiencing with the one-megabit chip. Shoji Tanaka of the University of Tokyo figures that the market for such devices will top $5 billion by the year 2000 and has enlisted the support of the Ministry of International Trade and Industry for a five-year national project to develop ULSI technology.
But the task may be difficult. Japanese manufacturers freely admit that much of the VLSI technology they have applied so successfully is imported -- particularly from the US -- and was developed in an environment that no longer exists.
US firms, thwarted from setting up wholly owned subsidiaries in Japan by a thicket of regulations finally discarded in 1975, frequently sold their technology to Japan under easy licensing terms. And in the late 1970s the International Trade Ministry sponsored a three-year VLSI project to bring Japan's semiconductor know-how up to that of the US.
Needless to say, the project succeeded. As a result, ``that sort of program isn't needed the way it once was,'' says Koichi Sumi, deputy director of the ministry's industrial-electronics division. ``And Japanese companies are being left to develop their own technology.'' Says Toshiba semiconductor chief Akio Kawanishi: ``No one's going to give it to us anymore.'' The investment payoff
Industry leaders are confident they are up to the challenge. Firms here are already strong in new materials research and development -- indispensable for the advent of new semiconductor technologies. Some observers say that the aggressive rates of capital investment practiced by many companies may prove to be even more significant in the long run. ``If you keep investing the way a lot of these companies have, you're bound to hit pay dirt,'' says James Abegglen, a consultant and professor at Sophia University here. ``If your capital rates are high, the level of innovation will move up.''
And so, Japanese executives say, it has been. Ten years ago Japanese researchers authored only a handful of the papers presented at the annual International Solid States Congress. At last month's meeting in New York, 49 out of 109 papers presented were produced by Japanese efforts. Japanese companies and government laboratories are generally considered to be doing about 70 percent of the world's research into gallium arsenide -- a material that promises startling improvements in speed and may replace silicon in the production of semiconductors -- and the US military has expressed interest in Japanese gallium arsenide technology.
Companies like Fujitsu are now pioneering gallium arsenide compounds. One example, called HEMT -- for high electron mobility transistor -- combines gallium arsenide and gallium aluminum arsenide to create a substance that could lead to circuits with operating speeds 25 times faster than silicon. Other companies have become world leaders in a process of making chips that consume little electric power; these are known as complementary metal oxide chips, or CMOS. Japanese expertise in this technology is, in part, a legacy of producing low power battery-operated consumer devices.
The Japanese have also begun to confound the experts with some success in semicustomized chips known as ``gate arrays.'' Fujitsu, for example, has become a leader in the field, and Dataquest estimates that Japanese worldwide gate array revenues will leap by more than 70 percent in 1985.
Indeed, the only area of development on which the jury is still out focuses on Japan's efforts to develop its own microprocessor technology. NEC, Fujitsu, Hitachi, and Toshiba are all working on state-of-the-art 32-bit microprocessor designs. But with the exception of NEC, they are opting to produce chips under license from market leaders like Intel and Motorola. NEC, which last fall began mass production of its new V series of microprocessors, will be marketing its own design by the end of 1985 -- the first original design microprocessor by the Japanese to be sold in the US. A patent infringement lawsuit file against NEC by Intel, however, has chilled the microprocessor climate in Japan somewhat.
``We're waiting to see how it turns out,'' says Dr. Kawanishi of Toshiba. ``What's the rush?''