What has Japan done to return to the ranks of semiconductor powers?

Japan's Ministry of Economy, Trade and Industry, industry and media eagerly hope that the chip factory set up by Rapidus in the Chitose area of ​​Hokkaido in 2027 will revive the semiconductor industry.

On September 1, 2023, Lapidus Chitose Factory officially laid the foundation stone. It plans to mass-produce 2-nanometer chips in 2027, with an investment of 5 trillion yen (approximately 250 billion yuan).

In 1988, Japan's semiconductor products accounted for 50.3% of the world's total market, but by 2006, only 25% remained, and in 2019 it even fell to 10%. The Ministry of Economy, Trade and Industry predicts that if the semiconductor industry is allowed to decline, Japan's share of the world's semiconductor products will be almost zero by 2030.

When the United States decided to suppress China in semiconductors, Japan immediately sensed an opportunity for a comeback.

The United States began to restrict companies’ semiconductor business dealings with Huawei in September 2019. In December of the same year, the Japanese government officially announced strict restrictions on semiconductor business dealings with Huawei and other Chinese companies.

Japan began to pay attention to the issue of economic security in 2020. In 2021 and throughout 2022, it established a set of economic security laws mainly targeting China from the legal level, cutting off ties with China in terms of cutting-edge semiconductor materials and semiconductor manufacturing equipment.

On the surface, Japan's economic security hinders the exchanges between Japanese companies and Chinese companies and damages the benefits obtained by Japanese companies in China. But looking at a deeper level, starting in 2021, the Japanese government will strengthen financial subsidies for the semiconductor industry. It is necessary to restore Japan's status as a semiconductor power as soon as possible by cutting off ties with China in cutting-edge products and technologies.

On August 10, 2022, Toyota, Denso, Sony Group, NTT, NEC, SoftBank, Kioxia, and Bank of Mitsubishi UFJ jointly invested 7.3 billion yen (approximately 360 million yuan) to establish Lapidus Corporation, which manufactures chips. (meaning "quick" in Latin), the Ministry of Economy, Trade and Industry immediately decided to provide the company with a subsidy of 330 billion yen (approximately 16.2 billion yuan).

On February 21, 2023, Lapidas decided to build a chip factory in Chitose, Hokkaido, using 7.3 billion yen to leverage an investment of 5 trillion yen (the ratio of registered capital to total investment is 0.146%). This may seem like a huge move, but more importantly, this company suddenly gave the Japanese semiconductor industry unlimited hope.

Why can a new company with only 7.3 billion yen give Japan great hope? In what aspects does Lapidus have a different business model from previous Japanese semiconductor companies? Can the Japanese semiconductor industry regain its glory because of Lapidus?

In the process of Japan's semiconductors turning from prosperity to decline, the Ministry of Economy, Trade and Industry has never stopped subsidizing semiconductors, but almost none of them has succeeded. Why is there hope of success this time? From Lapidus, we can not only see the characteristics of Japan's semiconductor policy, but also foresee the future of this country's semiconductor industry.

To make a big difference requires the support of public opinion

Judging from Japanese public opinion, Lapidus is different from previous Japanese semiconductor companies and has a quite new meaning.

In the past thirty years, Japanese semiconductor companies have basically taken the path of corporate restructuring when they encountered technical difficulties and poor management. But Lapidas is different. The company is a manufacturing company, a software company, and a semiconductor professional. It is a new enterprise jointly invested and established by banks and supported by the state.

In terms of investors, Toyota and Denso are the strongest manufacturing companies in Japan;

Sony, NEC and Kioxia, which became independent from Toshiba's most profitable flash memory division, are proficient in the manufacturing of semiconductor products;

SoftBank is an investor in ARM, the world's most important semiconductor circuit design company, and is an IT investment expert;

NTT is the most important data communications company in Japan, with huge demand for storage, computing, etc.;

Mitsubishi UFJ Bank's ability to raise capital is unparalleled in Japan.

In terms of manufacturing technology, familiarity with semiconductors, semiconductor users, funding, etc., Lapidas is the best in Japan. Although the amount of investment itself is not large, the strategic implications of making a big difference are significant.

However, having said that, Toyota earned 1.1209 trillion yen (approximately RMB 55 billion) in operating benefits in the second quarter from April to June 2023, but it lacks future products such as electric vehicles in important markets such as China. Fuel vehicles, including hybrids, will be available until 2035 at most.

When the company needs to transform, the capital invested by Toyota in Lapidus is only 1 billion yen, which means it is more asking for answers. For cars, including electric cars, it is good to use 48-nanometer chips, and 2-nanometer products will not be needed for a while. Investing in overly advanced chip companies is not part of Toyota's transformation.

Other companies, whether SoftBank or NEC, have suffered losses in recent years, preventing them from investing more capital. The 300 million yen invested by Mitsubishi UFJ is somewhat casual and not what the company really wants to do.

Japanese companies are dubious about the semiconductor policy implemented by the Ministry of Economy, Trade and Industry. Amari Akira, a politician who plays a high-profile role in semiconductor policy, once served as Minister of Economy, Trade and Industry. On November 11, 2022, Amari told Japanese reporters: "All past semiconductor strategies have failed."

This is true. After looking through relevant information on Japan's national semiconductor projects, I found that in the 10 years from 2001 to 2010, the Japanese government implemented Asuka Project 1 and Asuka Project 2, which ended in failure in 2010. The Milai project, which was launched at basically the same time, also failed to achieve any results. There was a Haruka project from 2001 to 2004, but it was stopped shortly after it was launched. The Tsukuba Semiconductor Cooperative, which was launched in 2006, barely survived until 2021 when it announced its dissolution.

After 2021, the Ministry of Economy, Trade and Industry first went all out to attract investment in Taiwan, China, and directly provided 600 billion yen in financial subsidies for TSMC to set up a factory in Kumamoto. But this is just to allow TSMC to bring 48-nanometer chip production technology to Japan and increase Japan's semiconductor production technology from three digits to 48 nanometers. This is not revolutionary in nature. Japan must have a company and chip factory established with domestic capital.

As for Lapidus, which only invested 7.3 billion yen in registered capital, it will eventually achieve a major project of 5 trillion yen. Companies such as Toyota have never said they will make additional investments and still need money from the state. After searching through the relevant information about Lapidas' financing, what we can finally get is still the hope that the state will provide money. The 330 billion yen will be provided in advance. After asking for advice, if there is not much opposition in Japan, the state will Further new subsidies will be provided.

On October 12, 2023, the "Asahi Shimbun" reported that the Japanese state invested another 1.7 trillion yen in Lapidas in addition to 330 billion yen, making up a subsidy of more than 2 trillion. However, this is still not the final amount. Most Japanese companies believe that it is a relatively high probability that the Japanese government will finally spend 5 trillion yen.

The Japanese state will not only increase subsidies to Lapidus, Japanese media reported that when TSMC sets up a second factory in Kumamoto in the future, the government will again subsidize 900 billion yen, and will also provide existing semiconductor products such as Sony's image sensors. An additional 700 billion yen in subsidies will be added.

What needs special attention is that the degree of subsidies for the localization of semiconductors in Europe and the United States is relatively limited. The subsidy rate in the United States is 5%-15%, and that in Germany is higher, reaching 50%. In contrast, the Ministry of Economy, Trade and Industry's subsidies to Japanese semiconductor companies are almost all-inclusive.

Although among Lapidus's investment companies, there are companies like Toyota that can earn more than 1 trillion yen in a quarter, the government will still heavily subsidize semiconductor companies, and the subsidies can span many years. Japan's private sector and relevant ministries and departments will not conduct strict audits on the results of subsidies. "The results will cause a lot of waste." "Asahi Shimbun" reported on October 12.

Therefore, it is not the great ability of Lapidas to make a big move, but more about persuading public opinion and getting the Japanese people to agree to allocate more resources to the country and foreign countries, even though they know that all previous semiconductor policies have failed. For investment companies and failures in the semiconductor business, individuals, companies, and governments will not be held accountable for any responsibility as in the past.

Lapidas' new idea

Judging from the information disclosed by Lapidus, the general development path before its establishment and in the future is as follows -

In August 2022, it was decided to invest 7.3 billion yen to establish the company. In November, it received a subsidy of 70 billion yen from the Ministry of Economy, Trade and Industry. In December, it established a strategic partnership with IBM and began dispatching to IBM research and development bases in 2023. Technical staff.

After these actions were approved by the Ministry of Economy, Trade and Industry, the government decided to add an additional 260 billion yen in subsidies in April. In September, it was decided to build the Chitose Factory in Hokkaido. In the future, the installation of manufacturing equipment will begin in December 2024, the trial production line will be completed in April 2025, and mass production will be officially launched in 2027.

It took a year and a half to build the trial production line, and after more than a year of debugging, it officially entered mass production. The efficiency of Lapidas should be quite fast.

From a technical point of view, the Chitose factory produces 2-nanometer chips, and TSMC's Kumamoto factory produces 48-nanometer chips. The two production technologies in Japan are not at the same level. Judging from the experience of setting up factories, TSMC not only has its own experience in Taiwan, China, but also built a factory in Arizona, the United States. Kumamoto is at least the third factory, and the products are relatively backward. The speed of building the factory is still not comparable to that of Lapidus. .

You must know that Japan proposed to build a 2-nanometer factory under the background that it has basically no experience in building factories above 48 nanometers, and the speed is so fast, which shows that it is quite confident.

This confidence probably comes from the fact that the main equipment is provided by domestic manufacturers in Japan. After the Dutch company ASML decided to cooperate with the United States in suppressing China's semiconductor industry, it did not have any important business on hand to assist Japan in quickly building Lapidus. .

The more important significance of Lapidas is that this company hopes to be different from Japan's previous vertically integrated production model and find a fast production model in addition to the horizontal division of labor model such as TSMC.

We can regard the previous Japanese semiconductor production and sales model as a "department store" model, which is vertically integrated. Whatever the company produces is sold in the market, and users do not have many opportunities to participate. Semiconductor companies can always come up with the latest products before users develop new products, and users use these new semiconductor products to design their own new products.

In other words, users can only do what they want to do behind the semiconductor manufacturers. Department stores have a sufficient supply of goods, and there is nothing that cannot be bought even if it is unaffordable.

However, manufacturers such as TSMC have changed the way semiconductors are produced, and foundry has turned manufacturers into huge customized factories. This is also called a horizontal division of labor production model. After the production method progressed to the horizontal division of labor, planning and design were handled by companies without factories (fabless), while production was handled by foundry companies.

Lapidus combines fabless companies and OEM companies responsible for design. Through early communication with GAFAM (Google, Amazon, Facebook, Apple, Microsoft) in the United States, after receiving relevant plans, it directly carries out the design and pre-processing. , post-process work. This shortens the manufacturing period and achieves the purpose of "quick" (the original meaning of the word Lapidas) production.

Three insurmountable mountains

Judging from public information, the Ministry of Economy, Trade and Industry has allocated a budget of 10 trillion yen in order to catch up with other countries and regions in the semiconductor industry, and a huge budget will be spent on Lapidus in the future. Lapidas has also not taken the initiative to raise funds, and is quite confident in getting subsidies from the state.

According to Japanese media reports on October 12, the Ministry of Economy, Trade and Industry has opened three funds in the production and research and development of semiconductors: "Research and Development Fund to Strengthen the Basics of 5G Information and Communication Systems", "Specific Semiconductor Fund" and "Ensuring Stable Supply" Support Fund".

Making semiconductors a top priority at the national level gives Lapidus every reason to use related funds provided by the government, and there is almost no need to struggle to raise capital like American and European companies, including external companies setting up semiconductor factories in Japan. In fact, now is the best time for enterprises to enjoy government subsidies and do what they want to do.

However, there are three insurmountable mountains in Lapidas.

First, there is basically no 2-nanometer chip market in Japan, while Lapidus has strong enemies such as TSMC and Samsung.

The biggest reason for the decline of Japan's semiconductor industry is that Japan's domestic demand for semiconductor products has declined sharply since 2000. The world's largest semiconductor market has moved to China. After Japan lost its domestic demand for semiconductor products, companies that produce semiconductors can only Despite the policy support of the Ministry of Economy, Trade and Industry and huge economic subsidies, the semiconductor industry was deteriorating. It was not until TSMC went to Japan to build a factory that the trend was reversed. However, the product market problem is still unresolved.

After 2020, the focus of world technological innovation has shifted to batteries, electric vehicles, 5G, industrial Internet, cutting-edge medical equipment, etc. Japanese companies control most of these technologies and patents, but Japanese companies have not failed to invest in the Japanese domestic market. There is no electric vehicle or 5G boom, and Japan does not have companies like American GAFAM or Chinese BATH (Baidu, Alibaba, Tencent, Huawei). The demand for storage and AI is very low, and there is no domestic market to build.

At the same time, economic security prevents Japan’s cutting-edge technologies and related products from being exported to China. Originally, semiconductor products could be exported to China to maintain the survival of Japan’s semiconductor industry. However, now 2-nanometer chips have to fully rely on the U.S. market. This uncertainty makes Japan's semiconductor industry increasingly needs to face new risks.

Lapidus' future 2-nanometer products will need to compete with chips from TSMC's factory in Taiwan, China, and the chip factory under construction in Arizona, the United States. TSMC is basically using depreciated production lines for production. The Arizona factory was built before Lapidus. If everything goes well, it will produce products before Lapidus.

In the case of a limited market, products from TSMC's Taiwan factory in China and products from the Arizona factory in the United States will also be supplied to GAFAM before Lapidus. Whether Lapidus' progress in production methods can cover TSMC's advantages depends on It is not technically possible to make accurate predictions.

Second, Japan is very short of semiconductor technology talents.

The construction of the Kumamoto factory basically uses up all the existing semiconductor talents. In the case of shortage of talents, Kumamoto Prefecture is actively building relevant training institutions to cultivate talents. Hokkaido is even more lacking in semiconductor talents. All Lapidas could think of was sharing talents with Kumamoto. If the factory officially starts operation, whether the method of sharing talents, especially the cooperation with TSMC, will be smooth, there are no relevant successful cases in the past, and it will be equally difficult to judge.

Lapidas is seeking cooperation with India. India has quite a lot of IT talents, but there are no ready-made semiconductor production personnel in India. Cooperation with India sounds very similar, but it is also very uncertain.

Third, Hokkaido’s water and power resources are limited.

A big reason why TSMC is optimistic about Kumamoto is its geographical advantages. Compared with Kumamoto, Chitose has insufficient water resources and only enough electricity for current consumption. Before Lapidas can set up a factory, Hokkaido needs to build a new water diversion canal for the factory and activate a new nuclear power plant. Especially in terms of the construction of nuclear power plants, Japan has experienced a long-term stagnation due to the accident at the Fukushima Nuclear Power Plant of Tokyo Electric Power Company. The administrative resistance to opening a nuclear power plant for a chip factory is quite large.

Even if the difficulties in terms of talent and geographical constraints can be overcome, in the end Lapidas will have to arduously move forward on a rugged road for a long time because there is no domestic market and competition abroad is too fierce.

Author: Japanese economic researcher.