Market Forces Compel Japan's Semiconductor Industry to Abandon the IDM Model

Yoshiko Hara, Editor-in-chief, SemiconPortal-EmergingTech from Japan

Recognizing that the traditional business model practiced by Japan's semiconductor companies is dead, industry players are compelled to reinvent themselves.

Japan's leading chipmaker Toshiba announced last year that it would outsource system-on-chip (SoC) large-scale integrations (LSIs) at 40 nanometers and beyond. Unveiled in the wake of the global financial crisis, Toshiba's restructuring plans revealed it was considering a fab-lite business model, but the company has yet to articulate a clear strategy.

Toshiba's announcement means no Japanese chipmaker intends to invest in advanced processes for logic devices. Moreover, it signals the end of the integrated device manufacturing (IDM) business model, long considered sacrosanct by Japanese semiconductor manufacturers.

In the 1980s, Japan's semiconductor industry captured almost half the worldwide semiconductor market. One legacy of that remarkable success is the conviction that outstanding quality sells products. The "soup-to-nuts" IDM business model—encompassing everything from development, design and production to marketing—has underpinned Japanese semiconductor manufacturers' quality assurance. But while the model may be good for quality, it is bad for profitability. Not only does the profitability of Japanese chipmakers trail that of several major overseas competitors, but Japan's share of the global semiconductor market has fallen for two decades, a decline that has accelerated since the middle of the last decade (Figure 1).

Figure 1. Japan's Shrinking Presence in the Global Semiconductor Market

Source: WSTS Inc., 2010

Source: WSTS Inc., 2010

Most semiconductor businesses in Japan were run as divisions of diversified electronics manufacturers. Then Hitachi, NEC and Mitsubishi spun off their semiconductor businesses. The dynamic random access memory (DRAM) business was the first to receive this treatment and Elpida Memory Inc. was founded. The treatment expanded to the logic business, with the merger and subsequent restructure of NEC Electronics and Renesas Technology last April. Now, like their competitors elsewhere in the world, Japan's semiconductor manufacturers are forced to stand on their own feet.

As processes shrink, compulsory facility investment soars. Thus, despite a company's desire to retain the IDM business model, keeping up with the industry's scaling trend requires an ample investment, which is unfeasible to maintain. Speaking last May, President and Chief Executive Officer Norio Sasaki of Toshiba Corporation noted that the IDM business model imposed such a heavy burden that Toshiba would have to shift to a fab-lite model.

Whereas several companies based in North America were early adopters of the fab-lite model at around the 90 nanometer generation, European chipmakers did so at the 90-65 nanometer generations. Now, Japanese companies are taking the plunge at around the 40 nanometer generation.

Japanese chipmakers' reluctant embrace of a fab-lite approach prompts the question: Can Japan keep its technological edge without advanced process fabrication?

In the recession triggered by the bursting of the "IT bubble" in 2000, an attempt was made to establish a "Japan foundry" so Japanese chipmakers could use advanced processes with minimum investment. This attempt fizzled out as the semiconductor market bounced back, restoring manufacturers' faith in their own facilities. Yet following a sense of crisis, a Japan foundry has once again become the talk of the industry.

Why Japan's Semiconductor Industry Must Change

New Competition

First, the electronics market has changed with the rise to prominence of the original design manufacturing (ODM) business model. The principal customers of Japanese semiconductor manufacturers have traditionally been Japanese manufacturers of electronic products, electrical gear and precision equipment. Based on their customers' demands, chipmakers developed high-performance custom LSIs with quality and performance often according priority over cost. This cozy world is threatened by the emergence of electronics manufacturing service (EMS) companies as major players in Taiwan and China. High-performance, high-cost devices do not attract these thrifty overseas manufacturers.

Two iconic consumer products, the Apple iPhone and iPad, are manufactured by EMS companies. The Japanese media frequently lament that teardown reports reveal that these popular Apple products contain few chips and components supplied by Japanese companies.

Cost-conscious Markets

In the past, Japanese manufacturers' obsession with quality gave the world personal computers containing DRAMs guaranteed for 25 years. But as many analysts point out, excessive quality entails unacceptably high cost in a cost-conscious market. And in these cost-sensitive markets, manufacturers must offer devices attuned to local needs at low cost. Traditionally, Japanese companies have developed high-end products first and low-end products as secondary versions. Furthermore, product introductions tend to be badly timed and local preferences ignored. Consequently, Japanese companies are behind their Korean counterparts whose grasp of local needs is demonstrated through astute marketing campaigns.

Disadvantaged Geography

Besides changes in the market, executives of Japan's semiconductor manufacturers often point out that companies manufacturing in Japan are at a disadvantage compared to overseas competitors. Some complain of what they view as inadequate government support for local manufacturers.

Last December, Prime Minister Naoto Kan directed ministers to prepare a 5 percent reduction in the effective corporate tax rate, from the current 40.6 percent to 35.65 percent. However, the tax rate is still high when compared to approximately 25 percent for China and Korea and about 30 percent across much of Europe.

Japan's Top Seven Semiconductor Companies' Business Models

Toshiba announced last December that it would outsource 40 nanometer and beyond SoC devices to multiple foundries. Toshiba owns two 300mm fabs, Nagasaki and Oita, for logic devices. The Nagasaki fab will be sold back to the original owner (Sony) while the Oita fab, which is migrating to the 45 nanometer process, will eventually concentrate on complimentary metal-oxide semiconductor (CMOS) sensors. For Flash memory production, Toshiba maintains an IDM business model and is continuing to invest in facilities in Yokkaichi to shrink processes, and 24 nanometer fabrication has already started.

Following a review and policymaking blitz in the first 100 days after the merger last April, Renesas announced in July that it would start outsourcing 40 nanometer devices and completely outsource fabrication to pure-play foundries for 28 nanometer devices and beyond. For 28 nanometer device fabrication, two foundries will be used, but eventually Renesas purportedly intends to integrate processes and use only one foundry.

Sony took the lead in announcing an asset-lite strategy back in February 2007. It is outsourcing production of 45 nanometer chips and beyond. Production of the 45 nanometer cell processor for the company's PlayStation 3 is outsourced to IBM. The Nagasaki 300mm fab, which Sony will regain from Toshiba, will concentrate on CMOS sensor fabrication.

Elpida, the sole DRAM maker in Japan, maintains an IDM business model, advancing the process generation with a relatively small amount of investment. It began mass production of 30 nanometer DRAM chips in January. Though a certain amount of DRAM production is outsourced to Taiwanese partners, Elpida is largely dependent on its own Hiroshima fab and subsidiary Rexchip in Taiwan for its DRAM capacity.

Panasonic announced the start of mass production of 32 nanometer Uniphier LSI chips last November. The company has partnered with Renesas for advanced process development. Their project, covering development up to the 28 nanometer generation, is almost complete. Beyond that, Panasonic has not disclosed its strategy. In May 2010, the company announced a three-year management strategy covering 2010 to 2012 without mentioning the semiconductor business.

ROHM maintains an IDM business model. It acquired Oki Semiconductor in October 2008. ROHM does not pursue cutting-edge processes but focuses on specific markets with devices one or two generations behind the scaling trend. It also concentrates on "more than Moore" technology, such as power chips fabricated on silicon carbide (SiC) wafers.

Fujitsu Semiconductor announced a fab-lite strategy in April 2009 and started outsourcing to Taiwan Semiconductor Manufacturing Company (TSMC) for 40 nanometer devices and beyond.

ASSPs and Emerging Markets Loom Large

How will Japanese chipmakers refocus their businesses as they embrace a fab-lite business model? Most are trumpeting a shift to application-specific standard products (ASSPs) and a reinforcement of overseas marketing.

Japanese semiconductor companies' main SoC LSI products have been application-specific ICs (ASICs), which are developed to comply with each customer's specifications. In contrast, ASSPs are standard products that semiconductor vendors propose based on marketing designed to reveal various users' needs.

In recent years, Japan's chipmakers have discussed the importance of ASSPs, but movement has been slow. Now, amid the competitive global market, they are seriously tackling ASSP development.

Renesas' target, for example, is the mobile communication LSI market. The company acquired a mobile communication modem unit from Nokia, and deploying the newly gained resources, established Renesas Mobile Electronics last December to develop and promote its LSIs for mobile phones. Renesas intends to be among the first to offer ASSP LSI chips for the coming long-term evolution (LTE) communication standard. The company expects mobile phone LSIs to be one of its growth engines and is challenging leadership in this field.

Toshiba intends to reverse the ratio of custom to standard product sales from the current 6:4 to 4:6 by 2012. It plans to promote an image processing engine, originally developed for the company's Regza liquid crystal display (LCD) TVs, as the core of its ASSP products. Consequently, the ratio of custom chips for game consoles and digital consumer products will decrease. The company formed a new marketing team last summer to push its ASSPs overseas.

Japanese SoC chipmakers have concluded that the vertical integration model is no longer sacrosanct. But because emerging markets are already crowded with competitors and overseas competitors are already leading in the ASSP market, prospects remain unclear. To gain ground, Japanese companies must deploy powerful marketing and show they have the capability to propose products that captivate local customers. Yet analysts and industry insiders suspect marketing may be a weak point because the marketing approach for the ASIC business is dramatically different than that required in the ASSP business.

Tsukuba Innovation Arena Has Global Ambitions

To support Japan's evolving semiconductor industry and spur growth, Tsukuba Innovation Arena (TIA), a research and development (R&D) umbrella organization, is taking shape in Tsukuba, Ibaraki Prefecture.

Two national research organizations—the National Institute of Advanced Industrial Science and Technology (AIST) and the National Institute for Materials Science (NIMS)—and Tsukuba University together with Nippon Keidanren, an industry lobby, announced in June 2009 their conception of TIA as a center for nanotechnology development involving the industrial, academic and government sectors.

In March, Japan's two big semiconductor-related projects, Millennium Research for Advanced Information Technology (MIRAI: meaning "future" in Japanese) and Semiconductor Leading Edge Technologies Inc. (Selete), an industry R&D consortium, will conclude.

MIRAI is a 10-year national R&D project launched in 2001 in response to the decline of the Japanese semiconductor industry during the 1990s. Selete is a consortium founded in 1996 by 10 semiconductor manufacturers to develop production technologies using 300mm wafer equipment. Since completing development for the 300mm wafer transition, Selete has played a major role in semiconductor R&D projects in Japan. TIA will take over some elements of these projects.

The TIA project is a national project funded mainly by the Ministry of Economy, Trade and Industry (METI). The distinction between the TIA project and conventional national projects is that the TIA project welcomes participation from overseas research institutes and manufacturers. Though national project policies do not explicitly prohibit overseas participation, practice has dissuaded it.

Tamotsu Nomaguchi, president of AIST, said that the organizers of TIA aim to foster TIA as a nanotechnology R&D base with global reach on par with the likes of IMEC, Albany and Minatech.

Figure 2. TIA Focuses on Six Core Research Domains Using Three Facilities

Source: TIA

With a new business model and a new R&D scheme, Japan's semiconductor industry is on the move. Japanese engineers remain confident that their technology has much to offer the world, yet the task of developing capabilities that correspond to market realities remains, and the market will not wait for Japanese vendors to complete this transformation. Japan's semiconductor companies have a tough fight ahead and will need strength and skill to stage a comeback.

About the Author

Yoshiko Hara is the editor-in-chief of SemiconPortal-EmergingTech from Japan. For more information please visit: www.semiconportal.com/en/. She can be reached at yoshiko.hara@semiconportal.com.

Back to Global Trends & Insights Home

Go to Global Trends