Can Germany Keep Taiwan in the Chips?


By Bruce Einhorn In the global semiconductor business, the ongoing saga of Micron's attempts to take over Hynix has taken center stage. But while the on-again, off-again talks get all the attention, Taiwanese chip executive Charles Kau has been quietly accumulating thousands of frequent-flier miles as he irons out a deal that could create shockwaves of its own. Kau, executive vice-president of memory chipmaker Nanya Technologies, has been the lead negotiator in talks with Infineon Technologies.

The German giant could hardly be more different than Nanya. Infineon has a storied history as a mover and shaker in the DRAM (dynamic random access memory) chip business, having been part of Siemens before being spun off as an independent company a few years ago. By comparison, Nanya is a newcomer: a 5-year-old upstart controlled by billionaire Y.C. Wang, who made his fortune in the petrochemical business as chairman of Formosa Plastics.

For all their differences, Infineon and Nanya now figure the rough-and-tumble of the memory-chip business is pulling them together. After six months of talks, the two in early May announced a tentative partnership. The deal, which they hope to close in a few weeks, will involve pooling resources for research and development, and forming a joint venture to build a spiffy chip fabricator in Taiwan. Cost of the plant: about $1.5 billion.

SURVIVAL TACTICS. With Japanese companies having largely exited the business and Hynix' fate uncertain, Kau figures a tie-up with a powerful patron is vital if Nanya is to have any hope in what he calls "this crazy DRAM industry," one of the world's most high-stakes and ruthless businesses. "If we don't do this," he asks, "how are we going to survive?"

The question of survival is one being pondered by a lot of memory chip executives all over the world. Having suffered through a disastrous 2001, the DRAM industry is in the final throes of a brutal consolidation. Just a handful of major players remains, down from some 20-odd companies in the mid-1990s. For those still around, the cost of survival is soaring, with companies facing the daunting prospect of committing billions of dollars to build the next-generation chip fabs that are increasingly becoming a necessity.

Players are getting little help from the notoriously volatile market for memory chips. Having staged a rapid recovery early this year amid hope Hynix would be the next to exit, prices have plunged once again, now that the Korean chipmaker's deal with Micron has collapsed. Due to poor market conditions, Nanya on May 31 announced it was postponing plans to sell more than 430 million shares overseas.

BROKEN MODEL? No wonder the worries are particularly strong among Taiwan's chipmakers. With the global shakeout having driven so many other companies out of the business, the Taiwanese are finding their business model under threat. Ever since they entered the memory chip business in the mid-1990s, most Taiwanese chipmakers have relied on others to provide them with vital expertise.

"All the patents for memory technology -- who owns them?" says Don Floyd, Asian semiconductor analyst for Lehman Brothers. "The Taiwanese don't. The Taiwanese have almost nothing." In the past, that wasn't a big worry, since the know-how was available from the likes of IBM and Toshiba, for a price.

Problem is, those companies have given up on the DRAM business, which leaves the Taiwanese alone -- and looking pretty vulnerable. Meanwhile, the cost of keeping up with the leading players is growing: New fabs that produce 12-inch wafers may be more efficient than the older 8-inch factories, but they're about twice as expensive, easily running over a $1 billion a factory.

"URGENCY." There's also the rising cost of technology. According to Nanya's Kau, companies need to spend about $300 million a year on R&D to have any hope of staying in the game. "The Taiwanese DRAM players have much more urgency in this situation than the European and American makers," says Floyd's London-based colleague, Ulrich Pelzer. "They don't know how these advanced memory chips are made, and so they need Infineon badly to have a future."

Small wonder Nanya has a lot of company in looking to Germany for a lifeline. Infineon recently agreed to cooperate with Winbond, an ailing Taiwanese memory chipmaker. The German company also is planning to increase its stake in ProMos Technologies, its joint venture with Mosel Vitelic, another Taiwanese semiconductor company.

The Taiwanese need the help. Despite the island's standing as an electronics powerhouse, with foundries like Taiwan Semiconductor Manufacturing and United Microelectronics -- the world leaders in the made-to-order chip business -- it is a latecomer in memory chips. Nanya, by far the biggest, has just 7% of the global market, vs. 25% for Micron and Samsung Electronics. Infineon is now No. 4, behind ailing Hynix, with about 14%.

BUILDING MUSCLE. By forming a German-Taiwanese axis, the companies are betting they can position themselves as alternative suppliers to PC giants like Dell and Hewlett-Packard, especially if Hynix' best assets get carved up and eaten by Micron. As part of an Infineon-led partnership, Kau says, they'll be among a group that commands a third of the market. And since computer markers are not likely to favor being stuck with just Micron and Samsung, according to this reckoning, the Taiwanese-Infineon united front can get some business. "We will have the muscle to stand up to Micron and Samsung," Kau boasts.

Of course, there's no guarantee the German-Taiwanese alliance will succeed. What happens if Hynix doesn't go away? Even if it does, how do the loosely affiliated Taiwanese partners of Infineon compete against Micron and Samsung -- while at the same time competing against each other? Alliance or no, there may be a lot more ugliness in the DRAM industry. Einhorn covers technology from Hong Kong for BusinessWeek. Follow his weekly Online Asia column, only on BusinessWeek Online


Best LBO Ever
LIMITED-TIME OFFER SUBSCRIBE NOW
 
blog comments powered by Disqus