It's merge or die for Japanese chip firms

It's merge or die for Japanese chip firms
TOKYO — A case of merger fever is spreading across Japan's ailing semiconductor industry, as some of its biggest players decide that it's better to join forces with competitors than risk being further marginalized in an unforgiving business climate.

The recent announcement that Hitachi Ltd. and Mitsubishi Electric Corp. intend to merge the lion's share of their semiconductor businesses underscores the sea change occurring in Japan. Until recently, it would have been considered unthinkable for two broad-based chip companies here to propose something so radical as combining their most valued products and technologies.

But now that Hitachi and Mitsubishi have agreed to combine the bulk of their chip business, there's talk that others will follow the same path. Two days after Hitachi and Mitsubishi announced their plan, Toshiba and Fujitsu — which in recent years have joined forces to develop core technology and to produce common networking DRAMs — were reported to be discussing ways to link their system-on-chip operations, according to the Japanese newspaper Nikkei Shinbun. The rumor had already been widely circulating in the analyst community, but last week Fujitsu officials in the United States called the report speculative and not based on fact.

Yet, some see further consolidation as inevitable. "As integration at this scale begins, it will stimulate other companies to merge their businesses. It's quite probable that midscale semiconductor companies will be integrated into major ones," said Yoshiharu Izumi, senior analyst at UBS Warburg (Japan) Ltd.

The writing has been on the wall for many years. Since the mid-90s, Japanese chip companies have been withdrawing from memory production either by pulling out of the commodity DRAM market, as in the case of Toshiba and Fujitsu, or by finding a partner, as NEC and Hitachi did when they jointly formed Elpida Memory several years back.

As the business climate has worsened this past year, Japanese chip companies have come under more pressure to put more of their core technology and devices out on the bargaining table — something they had resisted doing for years. But some say it's inevitable because so many large chip vendors in Japan sell essentially the same technology to a domestic electronics market that has been in a slow but steady decline.

"Too many semiconductor players are doing similar business in Japan. This is an obstacle to strong competitiveness in the world market," said Hitoshi Shin, first vice president of Merrill Lynch Japan. "Furthermore, these companies are running finished products along with semiconductors, which slows their decisions."

The debilitating semiconductor recession further served to prod Hitachi and Mitsubishi to merge their IC businesses. The semiconductor categories that will be shifted to the new joint-venture business generated $4.6 billion for Hitachi last fiscal year but should fall to $3.3 billion for the current fiscal year, ending this month. For the same areas, Mitsubishi has seen its sales go from $2.5 billion to $2 billion over the same period.

While both chip operations are in the red, the companies said they expect to be profitable in a year's time as the joint venture is put in place.

The new company will cover all non-memory products, including microcontrollers, logic ICs, analog circuits and discrete devices. The spin-off will also take over development of complex system-on-chip devices from its parents.

The venture, which will be equally owned by Hitachi and Mitsubishi, will have sales of $6 billion, making it the world's third-largest chip company, after Intel Corp. and Toshiba Corp., based on 2001 semiconductor market share figures from Gartner Dataquest.

Another round

This isn't Japan's first bout of merger mania. In 1999, Hitachi and NEC took the plunge to form Elpida Memory, a 50-50 joint venture for DRAM. Still, Japan's chip executives continued to exalt their home-grown logic and system-on-chip technologies as their competitive advantage.

But the strategy has largely fallen flat because of the enormous cost it entails. Many Japanese companies aggressively pursued embedded DRAM in the mid- to late-90s, but the technology required costly extra mask layers in the fabrication process and compromised logic performance enough that it never went mainstream.

Systems-on-chip, too, have been a disappointment because they are difficult to engineer and are used primarily in consumer applications like DVDs, where the competition is fierce. Some market analysts are uncertain whether combining the two companies' chip divisions will give them any advantage in the SoC segment. "I can't clearly see a bright scenario yet, because so few make a profit from the system-on-chip business," said Michito Kimura, a senior analyst at IDC Japan.

While many financial analysts consider mergers inevitable, some questioned the way this one is being executed. Saturo Rick Oyama, senior vice president of equity research at Lehman Brothers Japan Inc., expressed doubt that the new company could react to the market quickly enough if each parent company owns a 50 percent share. Elpida, he noted, has encountered that problem because all decisions have to be approved by both Hitachi and NEC. And the 50-50 arrangement's drag on the decision-making process could be felt even more acutely in the SoC segment.

"It can be fatal in the custom device business, such as system-on-chip LSIs, if a company cannot make a quick decision," Oyama said.

But Hitachi president Etsuhiko Shoyama rejected that reasoning. "Elpida is going well. There is no problem in decision-making," he said.

Saturo Ito, president and chief executive officer of Hitachi Semiconductor and Integrated Circuits, also downplayed the concerns. "The strategy is to take the lead in technology, not to focus on high-volume production, which may give people the impression that it has some problem," he said.

As for products, the Hitachi-Mitsubishi proposal still leaves some questions unanswered. It's unclear how the two companies will reconcile some overlapping product lines. Both companies, for example, are considered leaders in 16-bit microcontrollers and have spent much of their efforts recently promoting their own 32-bit devices. It appears the companies are already considering a common road map.

"It's desirable to integrate the architectures in coming generations. But for the time being, we can make use of each company's MPU for the most suitable applications," said Koichi Nagasawa, semiconductor group president at Mitsubishi.

It's also uncertain how the joint venture will affect SuperH Inc., the U.S.-based company that was spun out of Hitachi and STMicroelectronics. SuperH Inc. recently launched its 64-bit SH-5 device, which is being promoted as an embedded CPU for systems-on-chip.

A spokesman for Hitachi in the United States said that SuperH will likely have some role in the proposed Hitachi-Mitsubishi joint venture but added that the details have yet to be worked out. Asked whether ST's shares might be bought out, the spokesman responded: "There's no plan yet to buy ST's share. We're having discussions right now."

Products retained

Other products will remain within the parent companies, such as SRAM, flash memory and power and discrete devices. For Hitachi, that amounts to 17 percent of its semiconductor revenue. Mitsubishi will keep 40 percent of its chip business in-house. That business includes DRAM products that are being produced by Powerchip Semiconductor Corp. and Vanguard International Semiconductor Corp. in Taiwan.

Another area that still needs to be hammered out is which fabrication facilities will be transferred to the new company, the most important being Trescenti Technologies Inc., a 300-mm fab owned by Hitachi. The fab had been co-owned by Taiwan's United Microelectronics Corp. until recently, when the joint venture dissolved and Hitachi bought UMC's stake.

Hitachi's desire to increase the utilization of the expensive new line and Mitsubishi's hesitancy to build its own 300-mm line probably helped draw the two companies together. But "there is still a question of whether microcontrollers from the two companies are enough to run the line in full swing," said analyst Kimura.