Here’s Why Japanese Firms Lead the Chip-Making Pack
At the end of a winding mountain road 100 miles up the coast from Tokyo, a small upstart company is showing the United States why the Japanese now dominate one of the world’s most cutthroat businesses: semiconductor memories.
Here, in only six years of existence, NMB Semiconductor has built the most modern and lowest-cost chip production operation in the world. While countless U.S. companies have let their memory businesses languish--unable to keep up with the industry’s quick technological pace, fast product cycles and high production costs--NMB is in good health. It gets its factories up and running quickly, it offers superior quality and it’s been willing and able to invest hundreds of millions of dollars to stay competitive.
And all this from a company started by a ball-bearing concern that had no previous experience in electronics, let alone chips.
“They are on the cutting edge of production technology,” says Darrel Whitten, director of research at Prudential-Bache Securities’ Tokyo office.
NMB’s success suggests Japan’s semiconductor sector has lived up to the worst fears of industry watchers in the United States. Japan’s unique combination of a large infrastructure of firms making semiconductor manufacturing equipment, a steady supply of production engineers and abundant, low-cost capital have given the nation a decided edge. Now, even a new Japanese entry like NMB can outcompete such veterans of Silicon Valley as Advanced Micro Devices and National Semiconductor in chip production.
The young memory maker’s success also suggests that Japan’s production advantage in chips is far more important than many industry experts realized. Experts have usually pointed to the size and diversity of Japanese chip makers as the key to their success in taking 51% of the world market in semiconductors, while U.S. firms’ share has progressively eroded to just 35% last year.
Combining state-of-the-art technology obtained from American firms with Japan’s superior production technology, NMB now dominates the high-speed memory business, a small niche within an industry that has virtually become a commodity-type business because of its high-volume production.
Although NMB and other chip firms face tougher times now as worldwide semiconductor demand weakens and prices plummet, it has been highly profitable, earning a handsome $61 million last year on $260 million in sales.
Such rapid success didn’t seem possible in 1984, when a ball-bearing firm called Minebea asked Takumi Tamura, then a vice president at Sanyo Electric, to start a semiconductor company. Most Japanese chip makers laughed then and predicted the new venture would quickly fail. Depth of technology and long experience, they insisted, were crucial to success in the field.
Tamura, however, was always something of a maverick. Seeing Minebea’s offer as a chance to show his stuff, Tamura, then 67, left Sanyo and built his first plant with $250 million in capital from Minebea. Tamura put a handsome dormitory with pool nearby, and his own home on a peak further away overlooking a verdant valley.
The key to NMB’s success starts with low cost production.
In the semiconductor business, a manufacturer’s cost is closely tied to yield--the number of good chips it can get from each wafer of silicon.
NMB claims it has reached yields of up to 80% to 90% on its large-volume dynamic random access memory chips (DRAMs). That means only 10% to 20% of the chips produced are defective. U.S. companies have typically been happy with yields of between 50% to 60%.
The difference in quality and consistency between the U.S. and Japanese chip makers is “frightening,” says James A. Bixby, chief executive officer of Brooktree Corp., a San Diego semiconductor company that contracts out its chip production to firms in both countries.
A major reason for that quality gap is that Japanese companies have invested heavily in recent years. Consequently, their facilities tend to be newer than that of their competitors.
“This is the most advanced factory in the world,” Tamura, now 72 and chief executive of NMB, says of the firm’s just-completed “Number 2” factory.
Tamura says the new plant has the world’s first completely automated production line. The entire production facility is encased in glass, with robots moving trays of wafers from one set of machines to the next for further processing. When the wafers are done, they are placed on a cart on a linear motor car and taken to another room where they undergo automatic testing.
Since no one enters the production room, except for maintenance once a week, there is less danger of introducing dust particles that boost defect rates.
Even maintenance workers walk through a separate glassed-in hallway so that they are only exposed to the “clean room” area during short periods of time. The production equipment is controlled by computer from yet another room.
The clean room will be so clean, NMB official Souji Miyawaki says, it will far surpass requirements for Class I clean rooms, requiring less than one particle in the air per cubic foot.
Japanese production engineers’ famed commitment to quality helps. “If we have a problem on the production line, everybody will work through the night until the problem is solved,” says Miyawaki.
Another NMB strength is quickness. Tamura himself supervised construction of NMB’s first factory, completing it in a record six months. The second factory, which will be capable of manufacturing the most advanced memory chips now being designed, took one year to build.
“A factory that takes a year to build in Japan might take twice that long in the U.S.,” says Kazunori Hayashi, associate director of the semiconductor group in the Japan office of Dataquest, a San Jose-based market research firm. “When you’re dealing with a leading edge design, being the first to market is critical.”
U.S. firms have frequently been first out with a new-generation chip, but slower to get it to market. Texas Instruments was the first to announce sample shipments of 16-megabit DRAMs, the next new generation of chips, following 4-megabit chips that now are beginning to enter full-scale production. And Texas Instruments has recently boasted of vast improvements in the speed it takes to build plants.
But in previous new generations of chips, it was the Japanese who generally moved to mass production first.
Building a chip factory is easier in Japan because that is where most chip-tooling companies are based. Heavy investments in semiconductor factories by Japanese companies in recent years has helped nurture a powerful sector supplying manufacturing equipment. Those equipment suppliers can work with NMB to customize machines, install them quickly and provide good servicing.
“In Japan you can just order a (fabrication) line; you can’t do that in the U.S. any more,” says one U.S. industry executive.
Low cost of capital is also important to NMB’s competitiveness. Miyawaki estimates the company’s cost of capital to be about 6%. An important factor in its low capital costs: Minebea backing, which assures borrowers that the company is unlikely to go bankrupt.
Although Texas Instruments has managed to shave its capital costs to 14% now from about 18% a few years ago, it’s still at a significant disadvantage, says Norman Neuriter, president of the firm’s Japan operation.
NMB has managed to keep its products at the leading edge by buying designs from small U.S. chip design firms such as Vitelec Corp. of San Jose and more recently, Ramtron Corp. of Colorado Springs, Colo.
“It is cheaper than having the research overhead,” Tamura says. “Of the thousands of engineers at companies like Toshiba and NEC, only two or three are good designers. The rest are like assistants.”
For marketing, NMB has signed an exclusive agreement to sell its products through the distribution network of Intel, a leading U.S. chip maker.
Times are tougher today. With memory prices down 30%, earnings of all chip makers have plummeted. NMB’s dependence on the more sluggish U.S. market has put it in a weak position against its Japanese counterparts.
NMB’s earnings this year are expected to be down 46% on a 15% decline in sales. Nevertheless, unlike many of its U.S. competitors, the company continues to make a profit in memories.
“When profits are down like this there is a sense of crisis, but we just have to work harder to reduce costs,” says Tamura. “We have nobody else to turn to.”
Unlike most Japanese semiconductor makers, which are part of firms that also make computers or other chip-consuming equipment, NMB does not have a guaranteed internal market of computer and telecommunications product groups ready to buy its chips.
Yet Tamura fully expects to make it through the tough times. He’s done it before.
In 1985, soon after completing its first factory, NMB found itself in the midst of a slump so severe that analysts later estimated the Japanese chip industry lost $4 billion, versus a $2-billion loss for their U.S. counterparts.
NMB itself spilled as much as $123 million in red ink. Tamura says he just pointed out to the firm’s investors that he was at the point he had predicted on his business plan, and they left him alone. After a few tough quarters, earnings shot up along with the stock price.
Tamura is confident, once again, that better times aren’t far away. And he plans to keep investing in expanded production.
What NMB has yet to figure out--and what even the largest chip makers are now pondering--is how to finance production of future generations of chips. Building a plant for the current generation of DRAMs is close to $300 million. The next generation chip is expected to cost nearly $1 billion to design and produce.
“Even we have trouble financing that amount,” says a senior executive for giant Japanese chip-maker NEC Corp., explaining the rationale behind the extensive tie-ups these days between large American and Japanese semiconductor companies. NEC has a technology-sharing agreement with American Telephone & Telegraph. Texas Instruments is getting its capital through a venture with Kobe Steel, while sharing technology with Hitachi.
Many argue that the high capital requirements will lead to a shake-out, leaving just the top four or five manufacturers. “We don’t see NMB as a survivor. We just don’t see NMB in the market,” says Texas Instruments’ Neuriter.
But if Texas Instruments, NEC and Toshiba can find partners to share research and production costs, NMB says it will find someone too. That’s Tamura’s next challenge.
