DRAM fab strategies in Asia
09/01/1997
DRAM fab strategies in Asia
Ed Korczynski, West Coast Editor, Solid State Technology
There are many adoptable business strategies for a company and/or a country that wants to get involved in a hot new technology. Most strategies involve the new (junior) organization enticing a current technology-leading (senior) company into forming a relationship that leads to the transfer of key intellectual properties. This dynamic is continued by junior companies trying to join the big leagues of IC manufacturing, and by senior companies interested in optimizing capacity utilization.
An established technology leader in high-tech manufacturing such as a high-volume DRAM fabrication company needs to invest in both product design and manufacturing technology. In a rapidly changing business such as semiconductor manufacturing, it is imperative that a company maintain its technology edge to preserve market share.
Product design costs continue to increase with the number of transistors and the circuit complexity in advancing generations of ICs. An IC manufacturer can gain significant market advantage if it can devote a higher percentage of revenues to product design. Thus, strategies to reduce direct expenses and increase profit margins can directly translate into more advanced, noncommodity designs. In a magnifying effect, higher profit margins tend to create noncommodity devices that result in even higher profit margins to continue the cycle.
Critical to this "leveraged margin" strategy is the ability to ensure that direct manufacturing expenses are devoted to the highest-margin dice only. Low-margin dice that are subject to commodity-pricing pressures, such as DRAMs, generally do not require the highest level of processing technology and can be subcontracted to a foundry (See "Moore`s Law extended: The return of cleverness," Solid State Technology, July 1997, p. 359). The company`s own fabs can then be dedicated to the production of more proprietary higher-margin chips, such as logic, ASICs, or embedded DRAMs.
Outsourced designs are often only one product generation behind the leading edge. Sending mask sets and process flows for these chips to a foundry means that the senior company may create a future competitor that starts out only one generation behind. A wide variety of agreements can be included in contracts to protect the interests of the senior company as it sends its proprietary intellectual property out into the world.
History
In the 1960s, US companies created the IC manufacturing industry and enjoyed virtually unchallenged world dominance through the 1970s. Japanese IC companies, though at first the junior companies in low-margin and foundry partnerships, rose to challenge the more senior US companies in the 1980s. By the latter half of the 1980s, Japan effectively owned the DRAM business. Japan`s outstanding success in IC production can be directly traced to early US manufacturing partnerships.
"The secret is in doing it right," commented G. Dan Hutcheson, president of VLSI Research. "Back in the 1960s, US companies took their best technology to Japan and created their own competitors, which they had to learn how to deal with by the 1980s."
Suffering from declining profits and market shares in the latter half of the 1980s, US companies needed to develop new design and production strategies. One strategy played out by US companies with portfolios of memory chip designs was outsourcing of DRAM production to Korean companies. In so doing, US companies committed their futures to non-DRAM products such as microprocessors, DSPs, and ASICs.
"The US objective in the 1980s was to create the Koreans as competitors to the Japanese so that the US wouldn`t have competition in logic. By creating the Korean DRAM manufacturers, it guaranteed price pressure on the Japanese that prevented them from having sufficient resources to compete in logic," said Hutcheson.
Korean companies were interested in gaining world-class design and production experience, and DRAMs are perhaps the simplest high-volume IC designs to manufacture. Even technology that is one generation old is not easily obtained, and large internal investments can advance the technology once a manufacturing process is established.
Japan invested heavily in the 1970s, Korea invested in the 1980s, and now Southeast Asian countries are investing in the 1990s. For the prior two cases, there was a 5-10 year lag between the time of investment and gain of substantial market share. It is realistic to anticipate a similar delay before the newest members of the world IC manufacturing community step up as major players. Government decisions to invest huge capital sums create the fertile grounds that allow seeded technologies to grow rapidly.
Current strategies - embedded DRAMs
Fast forward 10 years to today, and the US strategy appears to have been a success. US companies dominate the high-margin logic segment, and Korean companies are formidable competitors to the Japanese in the memory market.
The next IC market battle looks like it will occur in embedded DRAMs. The market for embedded memory, defined
as an ASIC incorporating one-transistor (1-T) DRAM structures, is expected to grow to US$7.5 billion by the year 2001 (Fig. 1). There seems to be solid market demand for embedded DRAMs, particularly in 3-D graphics processing, digital set-top boxes, and other mixed signal products such as DVD and hard-disk drives.
|
Figure 1. Embedded DRAM forecast through the year 2001.
Some of the interest in embedded DRAMs can be seen as supplier driven, not market-driven. Semiconductor revenue fell substantially last year at all companies with a large DRAM percentage in their product mix (Fig. 2). Consequently, DRAM manufacturers are trying to increase revenues by finding higher-margin designs. The simplest design transition is from a pure DRAM to a still-partially-a-DRAM. Embedded memory proposals may, in part, be seen as solutions looking for problems to solve.
|
Figure 2. Worldwide DRAM revenue, actual and forecast, from 1994 to the first half of 1998.
Since Japan and Korea are the world leaders in DRAM production, it is no surprise that most IC companies in both countries have announced plans to develop and manufacture embedded DRAMs. Japanese IC companies look like they are attempting to apply the leveraged margin strategy against their Korean competitors, using junior companies in Southeast Asia in the same manner that the US used Korea against them.
"This same strategy is now being used by Japan against Korea," explained VLSI`s Hutcheson. "By shifting 16-Mbit DRAM production to Taiwan and China, Japanese manufacturers look to create competitors for Korea in the low end of memory. The Koreans are trying to enter into embedded memory, and the Japanese are trying to keep them out."
Taiwan, Singapore, and China have all been chosen for Japanese outsourcing. By seeding DRAM technology to rapidly evolving Southeast Asian companies, Japanese companies hope that profit margins will remain low at Korean companies. Forced to defend their high-volume production, Korean companies would then have a much harder time allocating sufficient resources to mount a viable challenge in newer products.
The Nihon Geizai, on April 21 of this year, reported that Fujitsu is planning to increase its outsourcing to 40% of its total production this year, using fabs in Taiwan and LG Semicon in Korea. The paper also indicated that outsourcing of memory chips was intended to free up capacity for new designs. Toshiba has reportedly entered into agreements with both Winbond Electronics Corp. (Taiwan) and Chartered Semiconductor Manufacturing Corp. (Singapore) for the outsourcing of not just DRAMs, but embedded DRAMs (Figs. 3, 4). Toshiba officials could not be reached for comment on the company`s outsourcing strategy.
|
Figure 3. Embedded 1-T DRAM buried trench structure.
|
Figure 4. Embedded DRAM roadmap showing bandwidth improvements from the integration of 1-T memory cells.
Development challenges
The greatest challenge to the success of the Japanese embedded DRAM production strategy may be the significant design and process integration investment required. The nonmemory component of these chips will require design efforts that are entirely different from those required by plain DRAMs.
"Once you know how to make a DRAM, you know how to make the next product," said Mel Thomsen of Pathfinder Research. "Once you know how to make 16 Mbit, you know how to make 64 Mbit, so your marketing is all internal. For logic, you need to get out and talk to customers to find out what the next generation product specs have to be, and while Japan has not been strong in this area, they`re trying to develop it."
The simplest way for an IC company to ensure healthy profit margins over the long term is to maintain a large portion of its product mix in intellectual, property-rich, noncommodity designs. Implicit in the production design wins of noncommodity chips are close working relationships with customers, consortia, and standards committees. Establishing these relationships requires adequate resources, time, and a compatible company culture.