Japan's semiconductor industry is betting everything on this factory

Japan is trying to create a semiconductor company that can catch up with giants like TSMC and Samsung in just four years.

In August 2022, eight Japanese companies including Toyota, Sony, Arm, and NEC jointly established the Japanese next-generation semiconductor national team, Rapidus. The Japanese government generously provided a subsidy of 330 billion yen.

Rapidus was born with the mission of revitalizing the Japanese semiconductor industry. The company aims to achieve domestic production of 2nm logic chips by 2027, on par with TSMC, and recreate the glory of Japan's semiconductor industry in the 1980s.

In September 2023, Rapidus officially started construction on its factory in Chitose City, Hokkaido.

The birth of Rapidus and its collaboration with IBM are considered Japan's "last and most important opportunity" to return to the cutting edge of semiconductor manufacturing. The bet on Japan's semiconductor industry is believed to be placed on this factory.

In the 1980s, Japan dominated the global semiconductor industry, but it fell into a slump after being suppressed by the United States for occupying the American market. Korean companies seized the opportunity and turned the tide. The previous mission to revitalize the Japanese semiconductor industry, Elpida, also retreated step by step under the pressure from Samsung until it went bankrupt.

Currently, Japan still has a dominant advantage in semiconductor materials and equipment, but can it catch up in the manufacturing process that has lagged behind for decades?

Can Rapidus replicate the myth of "Nihon Maru Semiconductor" or repeat the tragedy of "Elpida's failure"?

The rise and fall of Japan's semiconductor industry

In the 1960s, with the support of the United States during the Cold War (transferring hundreds of technologies including transistors, black and white TVs, and recorders to Japan), the unique craftsmanship of the Japanese people, and a large amount of cheap labor at that time, Japan's manufacturing industry began to flourish internationally.

However, as semiconductor technology transitioned from transistors to integrated circuits, Japan's semiconductor companies still lagged behind the United States by a whole generation.

In order to catch up with the United States, the Japanese government adopted the strategy of "trading market for technology" and mobilized the entire country's system to lead the "VLSI Joint Research and Development Program" from 1976 to 1979.

This "government-industry-academia" model eventually proved to be extremely efficient. Japan's semiconductor industry made rapid progress, and its DRAM chips not only had outstanding performance but also were priced 10% lower than those from the United States, quickly capturing the market.

The Japanese electronics industry in the 1980s was brilliant:

In the upstream materials and equipment sector, there were Tokyo Ohka and JSR's photoresist, and Nikon's dominant position in photolithography machines. In the midstream, DRAM held half of the global market share, all developed, manufactured, and tested by themselves, representing a true "fully independent industry chain," causing 80% of American DRAM companies to go bankrupt. At that time, the top five players in the DRAM market were the Japanese giants: NEC, Hitachi, Toshiba, Fujitsu, and Mitsubishi Electric. Intel, on the other hand, was on the verge of bankruptcy and was forced to withdraw from the memory chip market, transitioning to CPU development.

In the downstream end products, Japan was also thriving at that time: Sharp's panels and Sony's televisions were both doing well. Walkman and Fujifilm dominated the world outside the home, and Sony's Trinitron was synonymous with high-end televisions worldwide. In 1994, Sony's color TV shipments reached a staggering 100 million units.

However, it didn't take long for the Americans to strike back. The few surviving American chip companies temporarily set aside their differences and formed the Semiconductor Industry Association (SIA), initiating a nationwide confrontation.

The SIA created momentum through the media, hyping up the threat of Japanese technology and fueling anti-Japanese sentiment among the public. They also lobbied Congress, claiming that Japan's comprehensive leadership in this field posed a serious threat to US national security.

Under pressure from the United States, Japan first signed the Plaza Accord in 1985, causing the yen to appreciate sharply. In 1986, they were forced to sign the unequal Japan-US Semiconductor Agreement, which required them to open up the semiconductor market and ensure that foreign companies obtained a 20% market share within 5 years. During this period, the US launched continuous "301 investigations" against Japan and imposed retaliatory tariffs on Japanese semiconductor products. In 1991, Japan and the US concluded the Second Semiconductor Agreement.

After several rounds of semiconductor agreements and punitive tariffs, Japan gradually lost its technological and cost advantages in the semiconductor industry.

What's more critical is that with the support of the United States, Samsung from South Korea quickly rose to prominence.

Just like the blood transfusion given to Japan back then, the United States provided full support to Samsung, not only in terms of technical assistance from American engineers, but also in terms of suppliers and market analysis.

In 1982, the South Korean government implemented a "Semiconductor Industry Promotion Plan" similar to Japan's VLSI plan. In 1983, Samsung established a semiconductor factory and, with the help of the United States, mastered the key technologies of 16K to 256K DRAM in just 3 years.

In order to establish advanced production lines like Japan, Samsung aggressively poached personnel and equipment from Japanese companies. They bombarded executives of Japanese semiconductor companies with sweet words and organized a nearly hundred-person Japanese advisory group at a high price.

Faced with Samsung's catch-up, Japan became anxious, and Japanese companies started a price war with Samsung by selling memory chips at half the cost of Samsung.

As the price war broke out, chip prices plummeted, and American companies couldn't withstand the pressure. As a result, the United States launched anti-dumping lawsuits against both Japan and Samsung. However, the United States imposed a 100% anti-dumping duty on Japanese companies, while only imposing a 0.74% duty on South Korea!

Just as the price war was intensifying, the Japanese bubble economy collapsed in the early 1990s.

The Japanese finally couldn't hold on any longer. In 1992, Samsung took the lead over Japan for the first time, introducing the world's first 64M DRAM product. In 1993, Toshiba's storage semiconductor production was surpassed by Samsung, causing them to fall from the top position in storage semiconductors. In 1996, Samsung developed the world's first 1GB DRAM.

Elpida - Japan's Last Stand in the Semiconductor Industry in the Late 1990s

In 1999, amidst the turmoil in the Japanese semiconductor industry, Japan integrated the storage chip businesses of Hitachi and NEC to form a new company called Elpida. In 2004, Mitsubishi Electric's DRAM business was also merged into Elpida.

Elpida, which means "hope" in Greek, was essentially Japan's last stand in the semiconductor industry. The merger between Hitachi, which had strong capabilities in new technology research and development, and NEC, which had strong production technology capabilities, aimed to create the world's most powerful DRAM memory manufacturer.

This was Japan's final hope in the semiconductor industry.

With the support of government funding and policies, Elpida's products were labeled as "Made in Japan" and were also known as the "Japanese semiconductor." At one point, Elpida held nearly 20% of the global DRAM market share.

However, unexpectedly, just two years after the merger, Elpida's market share in the DRAM market plummeted from 17% to 4%.

According to analysts, the decline of Elpida actually began before its establishment. The Japanese semiconductor industry had already lost its competitiveness, and with American companies choosing to cooperate with South Korea, Japan was isolated. The establishment of Elpida was nothing more than a last-ditch effort.

Traditionally, Japanese semiconductor companies focused on solving problems at the root by improving the design and manufacturing processes, with the premise of not producing defective products. This approach not only improved product reliability and yield rates but also ultimately increased production efficiency, allowing Japanese semiconductor companies to continuously output low-cost DRAM products.

However, with the advent of the PC era, the importance of Japan's proud yield rates and reliability diminished. While mainframe computers may require DRAM that can last for 25 years and consider sustainability and stability, personal computers are often replaced every 5 years, emphasizing rapid updates and cost control.

But even until the establishment of Elpida, the Japanese never gave up their pursuit of high-quality products. Elpida also exposed the drawbacks of a restructured company: Hitachi excelled in new technology research and development, NEC excelled in high yield production, and Mitsubishi excelled in integration and coordination. In the power struggle, NEC eventually gained the upper hand and remained complacent in the cycle of being the "technology leader" in the Japanese semiconductor industry, with no interest in low-cost production technology for DRAM.

In order to achieve high yield rates for DRAM, Elpida continued the expensive equipment usage habits of Japanese semiconductor companies. Each process was equipped with dedicated equipment, resulting in Elpida's wafer throughput being only half of Samsung's, even if Elpida had a higher yield rate, it still couldn't match Samsung Electronics, which had twice the yield rate.

As a result, Elpida, with its higher technological level, only had a profit margin of 3%, while Samsung Electronics had a profit margin as high as 30%. Therefore, Samsung's DRAM, which was considered a knockoff product by the Japanese, rapidly expanded its market share based on its cost-effectiveness, riding the wave of the booming PC market. On the other hand, when Dolphin Research was established, the chip industry had already shown a trend of separating the development of underlying IP, design, and manufacturing, which Japanese companies failed to keep up with.

As Moore's Law steadily advanced, only companies like Samsung and Intel were able to rely on the market share of downstream end products to support their technological investments in manufacturing. Japanese IDM (Integrated Device Manufacturer) companies, which lost their positions in PC and smartphone terminals, found it increasingly difficult to bear the massive investments in both manufacturing and design, ultimately falling behind in both areas.

Around 2008, due to the sluggish sales of Windows Vista, the oversupply of DRAM led to a drastic drop in prices. Then, the financial crisis hit, causing a downturn in the semiconductor market demand. Dolphin Research watched helplessly as DRAM prices fell below $1.

At this time, Samsung, which was already a giant in the industry, aggressively expanded production during the period of falling DRAM prices and excess capacity at Dolphin Research, squeezing Dolphin Research's market share despite losses. Accumulating massive losses and debts, Dolphin Research retreated step by step, even with the strong support of the Japanese government, it was still unable to turn the tide.

In 2012, Dolphin Research went bankrupt and was acquired by Micron for a mere $2 billion.

Thus, the last hope for storage chips in Japan was extinguished, and a pillar industry collapsed in Japan, leaving behind a historical feud between two East Asian countries.

A Decade of Dormancy

With Dolphin Research's bankruptcy, the entire Japanese semiconductor industry collapsed. Panasonic, Sony, and Sharp, the three major giants, suffered record-breaking losses, and Renesas Electronics (established in 2003 through the merger of Hitachi and Mitsubishi Electric's semiconductor divisions) teetered on the brink of bankruptcy.

The bankruptcy caused a significant shockwave and brought far-reaching secondary disasters to the Japanese industry:

First, the decline of terminal brands. Sharp's TVs, Toshiba's air conditioners, Panasonic's washing machines, and Sony's smartphones all struggled, and consumer electronics giants almost all shrank into component suppliers. Sony was hit the hardest, as its advantageous projects such as cameras, Walkmans, and audio-visual systems were all outcompeted by the iPhone one after another.

Second, the collapse of the upstream supply chain. From panels and memory to chip manufacturing, Japan lost almost every battle against the Koreans. The once dominant Japanese storage chips were reduced to only Toshiba's flash memory. However, due to Toshiba's struggles in transitioning to nuclear power and the impact of financial fraud, the flash memory business was renamed Kioxia and tearfully sold to Bain Capital.

In the following decade, Japanese semiconductor companies adopted a "backing support" strategy, which means avoiding the "glamorous" finished product market and focusing on the "back-end market" such as equipment and materials. At the same time, companies like Renesas and Sony gradually narrowed their focus to their areas of expertise.

Renesas completely withdrew from the smartphone market and refocused on its traditional core business, MCUs (Microcontrollers). MCUs, commonly known as single-chip microcomputers, have the largest application in the automotive industry. Automotive MCUs have always been Renesas' most profitable and advantageous business, occupying nearly 40% of the global market. Sony, on the other hand, has invested its limited R&D resources in the digital imaging business, represented by CIS chips, and has participated in the mobile terminal wave as a component supplier.

CIS chips are electronic devices that convert optical images into electrical signals. They are essential components in smartphones, commonly known as "image sensors." In 2011, Sony's IMX145 was first used in the iPhone 4s, and the concept of CIS chips became popular. Since then, from Samsung's S7 series to Huawei's P8 and P9 series, Sony's CIS chips have become almost standard in flagship models.

Starting in 2016, Sony gradually shifted its strategic focus on CIS chips to the automotive field. CIS chips are responsible for capturing environmental information in autonomous driving, acting as the "eyes" of the car, and their importance is self-evident. By 2017, Sony unveiled its three-layer stacked CMOS image sensor at the ISSCC conference, solidifying its dominant position.

Unlike chips like CPUs and GPUs that rely on increased integration for computational power, MCUs and CIS chips, as "functional chips," do not require high advanced processes but have higher requirements for reliability and durability. They heavily rely on craftsmanship.

Compared to Sony's high-end CIS chips, which still require TSMC's foundry services, Renesas' MCU products mostly use 90nm or even 110nm process technology. The technological threshold is not high, and the pace of updates is slow, but the lifecycle is long, and once customers choose them, they are unlikely to switch easily.

Therefore, although Japan's memory chips have been severely challenged by South Korea, Japan has almost never been left behind in the industry's dominance represented by analog chips.

Once known for its belief in technology supremacy, Japan achieved semiconductor glory in the 1980s. However, due to its reliance on this belief, it later experienced a decline and the bankruptcy of companies like Elpida. But in the end, it was this belief that allowed Japan to preserve the spark in the semiconductor industry, from MCUs and CIS chips to semiconductor materials and equipment.

However, compared to the panel and memory sectors, these fields are not considered large markets. This is also the reason why the Japanese government and industry find it difficult to be satisfied: compared to the traditional advantageous projects they have lost, what they have gained is too little.

According to IC Insights data, in 2021, the global semiconductor market share was 54% for the United States, 22% for South Korea, 9% for Taiwan, and only 6% for Japan.

Therefore, the Japanese government has never given up its goal of regaining lost ground. This is also the important background for the establishment of Rapidus, which aims at the 2nm process.

Rapidus - "The Last and Most Important Opportunity"

In the past thirty years, the suppression from the United States has left Japan with a psychological shadow and lost confidence in making a comeback.

However, today, with the changing times, companies like TSMC and Samsung have fallen out of favor, and Japan is once again partnering with the United States in the semiconductor field. Last year, Japan announced a joint investment of 1.3 trillion yen with the United States for the development of the next-generation semiconductor and established the national team Rapidus. In November 2022, Rapidus announced a partnership with IBM in the United States, and the following month, they reached a collaboration with IBM on the transfer of IBM's 2nm technology.

In 2021, IBM became the first in the world to successfully produce chips using the 2nm process. It has been reported that Rapidus has already dispatched 100 technical personnel to IBM to learn the GAA (Gate-All-Around) technology required for the 2nm process.

In May of last year, the then Japanese Minister of Economy, Trade and Industry, Koichi Hagiuda, expressed his thoughts during his visit to the United States:

"In cooperating with the United States in the semiconductor field, I feel the strangeness of fate."

The Japanese government sees the current situation as the "last and most important opportunity" to revitalize the Japanese semiconductor industry.

Over the past thirty years, the Japanese semiconductor manufacturing industry has been in decline, especially in the field of logic semiconductors. Japan lacks the capability to design and develop advanced logic semiconductors and its production is still at the 40nm process node and above. The production capacity for the 40nm-90nm process node accounts for only 18%. The Japanese industry believes that the Japanese semiconductor manufacturing industry is facing a critical period of life or death.

In addition, although Japan still has strong competitiveness in upstream semiconductor materials (with a market share of about 50%) and equipment (with a market share of about 30%), it also faces competition from its East Asian neighbors and cannot be complacent. This is also true for the last barriers such as MCUs and CIS.

Some analysts in Japan believe that there is a sense of urgency between the Japanese government and industry insiders. The current situation is seen as the "last opportunity" for the revitalization of the Japanese semiconductor industry, especially the manufacturing industry. If this opportunity is missed, the Japanese semiconductor manufacturing industry may completely lose its international competitiveness by 2030.

Furthermore, Japanese semiconductor experts believe that the semiconductor industry is no longer simply following Moore's Law. It is evolving from the "Deep Moore" (continuing to advance in the planar direction along Moore's Law) that started late in Japan to the "Beyond Moore" (crossing Moore's Law and seeking development through advanced packaging technologies such as stacking and chiplets) route, such as 3D packaging. With its competitiveness in equipment and materials, the Japanese semiconductor industry has the opportunity for a resurgence.

With the rapid development of the digital industry and the increasing importance of semiconductor technology in its development, Japan hopes to win the competition in fields such as data centers, autonomous driving, the next-generation communication standard 6G, and metaverse. Therefore, it is urgent to reshape the industry chain, develop advanced chip processes, and seize the commanding heights of future digital technology competition.

Replicating the Past Model

In the past two years, Japan has been making frequent moves in the semiconductor field.

In March 2021, the Ministry of Economy, Trade and Industry of Japan established the "Semiconductor and Digital Industry Strategic Research Association".

In June 2021, the Ministry of Economy, Trade and Industry announced the "Semiconductor and Digital Industry Strategy", which outlined a "three-step" strategy - restoring semiconductor production capacity, promoting the development of next-generation semiconductors, and laying the foundation for future technologies. The first important step is to introduce TSMC. The Japanese government provides a subsidy of approximately 476 billion yen, inviting TSMC to build a semiconductor factory in Kumamoto Prefecture. It is reported that the new factory in Kumamoto will start operation in December 2024.

The second step is to construct the mass production system of the next-generation semiconductor using the "dual-drive" approach. One is the establishment of a joint venture called Rapidus by eight Japanese companies in August 2022 to undertake semiconductor chip production business. The other is the establishment of the "Leading-edge Semiconductor Technology Center (LSTC)" led by the Ministry of Economy, Trade and Industry in November 2022, responsible for establishing an open research and development team.

It still tastes familiar, like the formula of "government-industry-academia" in the 1970s.

The final third step is that Japan plans to change the rules of the game after 2030 by developing optical-electrical fusion technology based on NTT's IOWN concept.

Can they recreate the myth of the past?

The Latin meaning of "Rapidus" is "fast", indicating that Japan will start a speed war.

The previous representative of "hope", Elpida, eventually shattered Japan's hope. Now, can Rapidus, representing "fast", enable Japan to achieve mass production of cutting-edge chips from 2027 onwards and reproduce the glory of the semiconductor industry in the 1980s?

According to analysis, overall, Japan already has the industrial foundation to develop advanced chip processes, and even in the core manufacturing process of semiconductors, it has certain competitiveness. At the same time, Japan chooses to strengthen cooperation with the United States in the semiconductor industry, giving it a significant advantage in materials, equipment, and design. It can be said that besides lithography equipment, the United States and Japan can almost produce and supply everything independently. Semiconductor materials and EDA design software are in an absolute monopoly position.

However, if Japan wants to fill the gap in advanced semiconductor manufacturing processes in the past 20 years within a few years, it still needs to face two major problems: talent and funding.

It is reported that the manufacturing process of 2nm semiconductors involves approximately 1,000 to 2,000 steps, generally requiring about 1,000 technical personnel. Currently, Rapidus has only recruited more than 200 employees, and the president and CEO of Rapidus, Atsuyoshi Koike, stated in an interview that most of the existing engineers are over 50 years old. However, Koike also stated that AI and automation technologies will be introduced in the manufacturing of 2nm semiconductors, which can reduce the manpower requirement to around 500 people.

In terms of funding, the investment in 2nm chip technology is astonishing. For example, TSMC's 3nm chip production line in the United States requires an investment of 40 billion US dollars. This is also the reason why Rapidus Chairman Tetsuro Higashi proposed a funding requirement of 54 billion US dollars (approximately 7 trillion yen). Previously, Junichi Koike told the media that Rapidus and TSMC have different business models and he is not worried about the talent shortage issue. However, he admitted that there is currently a lack of funds, and one of the solutions is an IPO.

In addition, there are also hidden concerns in terms of technology. Japan still lags behind China's Taiwan and South Korea in semiconductor manufacturing. TSMC and Samsung have already started mass production of 3nm products and plan to start mass production of 2nm products as early as 2025. Currently, Japan's main semiconductor production lines are still at the 40nm stage. In order to upgrade the semiconductor manufacturing industry, they are also attracting TSMC to build a factory in Kumamoto Prefecture to produce 12-28nm logic chips.

Therefore, it can be said that Rapidus, starting from scratch, directly launching the 2nm chip process, is it a "technological leap forward"? Even if Rapidus has IBM's support in 2nm chip technology and builds a 2nm production line, it will still lag far behind TSMC and Samsung. Moreover, it is very likely to become a fatal key to commercial application, just like Samsung's low yield rate in the 3nm chip process.

After IBM sold its semiconductor division to GlobalFoundries in 2014, the company announced its withdrawal from chip foundry business and only focuses on chip technology research and design. This means that IBM cannot have the same rich practical experience as TSMC, which has been in the forefront of manufacturing, and its ability to implement technology is a big question mark.

In addition to technology, talent, and funding, even if Rapidus really achieves a miracle in a few years, solves the yield rate issue, it still faces the problem of how to acquire domestic and foreign customers.

In the field of semiconductor foundry, TSMC has more than half of the global market share, and Samsung Electronics and Intel from the United States are also involved in foundry business. Rapidus has no chance of winning against these giant companies.

The most important thing in wafer foundry is economies of scale. Without end customers like Apple who purchase advanced process chips on a large scale, it is doubtful whether the revenue generated from relying on server and supercomputer customers can bring profitability, let alone guarantee the subsequent process iteration.

Moreover, the United States has implemented policies to support the chip industry, intending to bring the global semiconductor supply chain back to the United States and elevate semiconductor technology competition to the level of national strategy. If IBM's 2nm technology can be successfully implemented, will the United States turn a blind eye to Rapidus, a non-domestic company taking on advanced chip processes?

Back then, after defeating Nippon Electric, the Koreans celebrated for a long time. They cheered for themselves for finally reaching the pinnacle of the world and defeating Japan.

The Koreans did not expect that they were not the ones laughing in the end. The United States never forgot about the Korean semiconductor companies.

As a result, Samsung now has 55% of its shares controlled by foreign capital, most of which are American.