Published: April 14, 2026
In the last volume, I mentioned that horizontal specialization has not necessarily advanced across all product fields. This time, I will discuss how horizontal specialization has not necessarily advanced in every country.
Horizontal Specialization Has Not Necessarily Advanced across All Countries
Countries Late to the Semiconductor Industry
(1) Taiwan
(2) South Korea
(3) China
Countries Early to the Semiconductor Industry
(1) United States of America
(2) Europe
(3) Japan
Did Japan’s Semiconductor Industry Decline Because It Could Not Cope with Horizontal Specialization?
Volume 18: Horizontal Specialization in the Semiconductor Industry—Horizontal Specialization by Product and Country (Part 2: By Country)
In the previous volume, I explored the degree of horizontal specialization by product. Now, let’s do the same for countries.
Even if horizontal specialization is advancing overall, that doesn’t mean it is advancing evenly within each country. As explained in Volumes 15 and 16, foundries and OSATs are predominantly concentrated in Asia. In addition, as I mentioned in Volume 13, the top fabless companies by revenue are based in the U.S. and Taiwan. Companies from three countries—the U.S., Taiwan, and China—account for nearly 100% of fabless companies’ share of revenue. Let’s look at each country below.
Note: According to IC Insights, the revenue share of fabless semiconductor companies in 2021 was 98% in the United States, Taiwan, and China combined (68% in the United States, 21% in Taiwan, and 9% in China), while Japan, South Korea, and Europe’s share was 1% or less. IC Insights, the company that conducted the survey, closed at the end of 2022 due to the retirement of its founder.
The degree of horizontal specialization varies from country to country
First, let’s look at countries that entered the semiconductor industry relatively late.
Note: This article is an English translation of the Japanese version written in late 2023. As such, the story is based on revenue ranking data from 2022 and 2023. However, even with slight variations in the data, the explanation in the following paragraphs will remain relevant as of May 2025.
Taiwan is the birthplace of foundries. According to TrendForce data, four Taiwanese companies, including Taiwan Semiconductor Manufacturing Company (TSMC), which holds the overwhelming No. 1 share, and UMC, a regular in the top five, were ranked in the top 10 for foundry revenue in the second quarter of 2023. Taiwanese companies account for about two-thirds of the total market share. Additionally, according to IDC data, six Taiwanese companies, including ASE, ranked in the top 10 for OSAT revenue in 2022. It appears to be the only country with an established system of horizontal specialization that allows all semiconductor manufacturing processes to be completed domestically. Fabless companies that use horizontal specialization have the second-largest revenue share after U.S. companies. As discussed in Volume 13, MediaTek was one of the top 10 semiconductor companies in terms of revenue in 2021 and is a Taiwanese company.
On the other hand, there are no IDMs ranked high in revenue, so one can say that the presence of IDMs in Taiwan is relatively weak. Although there are IDMs that manufacture memory, such as DRAM, their market share is small.
South Korea, like Taiwan, entered the semiconductor industry relatively late. Nevertheless, its two leading semiconductor companies—Samsung and SK Hynix, both IDMs—consistently rank among the top 10 by revenue. Samsung is an IDM that also runs a foundry business, ranking second in foundry revenue. DB HiTek is also a leading foundry that is sometimes ranked in the top 10 in terms of revenue. However, the absence of high-revenue fabless companies suggests South Korea plays only a limited role in the global landscape of horizontal specialization. The country is significantly populated by two major IDMs, Samsung and SK Hynix. On the other hand, the South Korean government is also making active efforts to foster fabless companies.
When I investigated it, I found that the history of China’s semiconductor industry is surprisingly long. However, full-scale development didn’t begin until the 1990s, so it seems reasonable to consider China a latecomer. The country initially aimed to nurture IDMs but shifted its policy toward nurturing foundries in the late 1990s. As a result, domestic foundries are growing. Three companies placed in the latter half of TrendForce’s 2023 Q2 top 10 foundry revenue ranking. In OSAT as well, three companies from China were included in IDC’s 2022 top 10 OSAT revenue ranking. In both foundry and OSAT, China comes right after Taiwan in terms of the number of companies listed. I think it is reasonable to say that a certain framework for horizontal specialization has been established domestically. However, because no domestic foundry offers leading-edge process technology, Chinese fabless firms must rely on overseas companies like TSMC. As of 2020, more than 2,000 fabless semiconductor companies reportedly relied on these foundries and OSATs. As noted above, their combined revenue share ranks third after the U.S. and Taiwan. Note that the gradual tightening of U.S. semiconductor export restrictions to China since 2020 has made it difficult for China to use cutting-edge semiconductor technology from outside the country. Consequently, China has been working to strengthen its domestic semiconductor development and manufacturing capabilities, in other words, to achieve self-reliance.
China is also aiming for domestic production of memory products such as DRAM and NAND flash. However, the companies that are attempting this are IDMs, not fabless companies. As mentioned in the previous section, IDMs play a leading role in manufacturing these products. Although there are other IDMs in existence, none of them are currently among the top-ranked in terms of revenue.
Next, let’s take a look at countries that entered the semiconductor industry early.
First, as mentioned in Volume 13, the U.S. has many leading fabless semiconductor companies. GlobalFoundries, which originated from AMD’s manufacturing division, is ranked among the top five in terms of revenues. However, fabless companies rely heavily on foundries and OSATs in Asia, especially in Taiwan, for their production. It can be said that it is not based on domestic specialization, but rather on international one. However, this does not mean that IDMs have a small presence. It is the birthplace of the semiconductor industry. There are traditional and powerful IDMs, such as Intel and Texas Instruments (TI), and three IDMs, including Micron, are among the top 10 in revenue. Strong fabless companies and strong IDMs coexist.
Note: Amkor is the second-ranked OSAT company in terms of revenue. Although it is a U.S. company, all of its manufacturing facilities are in Asia. There are manufacturing sites in Japan, but they used to be back-end process plants of Japanese IDMs. As I mentioned in Volume 16, the company was originally the U.S. base of a South Korean company called ANAM. However, in December 2023, the company announced plans to build an advanced packaging factory in the United States.
Unlike in the U.S., the leading semiconductor companies with the highest revenues in Europe are IDMs. Three European companies ranked 12th to 14th in Omdia’s 2022 top 20 revenue ranking. All three are IDMs. As for fabless companies, the share of revenue for European fabless companies was less than 1% in 2021, down from 4% in 2010 (data from IC Insights). This appears to be due to the acquisition of a leading fabless company. Although there is a leading foundry, X-FAB, Europe’s presence in the world of horizontal specialization is considerably smaller than that of the U.S., Taiwan, and China in terms of revenue scale.
Note: The leading European fabless companies that were acquired include: Dialog and CSR were among the top 15 companies in the revenue ranking. Dialog was acquired by Renesas in 2021 and CSR by Qualcomm in 2015. Both companies are British. Panthronics, which Renesas announced the acquisition of in 2023, is an Austrian company, and Sequans is a French fabless company. However, in February 2024, Renesas announced the cancellation of the Sequans acquisition.
However, I would like to introduce a fabless company in Europe that plays a leading role behind the scenes. This company is Arm, a British firm. You may know the company from the news of its acquisition by SoftBank in 2016 or its listing on NASDAQ on September 14, 2023. Although it does not manufacture its own products, it is a fabless company that provides processor technology. Many products contain ARM-based processors. For example, most smartphone system on a chip (SoC) devices contain them.
Similar to Europe, all of the leading semiconductor companies in Japan are IDMs. Although fabless companies exist, they are few in number and have little global presence. In August 2022, the establishment of Rapidus was announced. Rapidus aims to become a cutting-edge fine process foundry. Currently, however, there is no cutting-edge foundry in Japan, so Japanese companies have to rely on those in other countries. Japan’s presence in horizontal specialization is even smaller than Europe’s, but there is hope for future developments. Notably, Rapidus has begun building a pilot line at its factory in Chitose, Hokkaido, as of April 1, 2025.
As mentioned above, the degree of horizontal specialization varies from country to country. I think it is safe to say that Taiwan is the only country with an established system of horizontal specialization within its borders and the ability to produce almost all semiconductor products using its own foundries and OSATs. Horizontal specialization in semiconductors is built upon global specialization. As mentioned earlier, production heavily relies on Asia, particularly Taiwan.
Note: Taiwan also cannot supply its own semiconductor manufacturing equipment, materials, and design tools, and relies on companies in the U.S., Japan, and Europe. In this sense, ultimately, no country can produce everything within its own borders.
However, the semiconductor shortage between 2021 and 2022 has triggered a growing awareness that semiconductors are an important strategic commodity. With rising geopolitical tensions, the extent to which semiconductor production depends on Asia, especially Taiwan, has become an issue from the perspective of economic security. There is a movement to bring production domestically. The establishment of Rapidus can be considered a related move. However, TSMC is currently the only pure-play foundry capable of producing cutting-edge fine processes. There is both a bidding war to attract TSMC and a race to offer subsidies as incentives. The first plant in Kumamoto began mass production in December 2024, and the second plant is scheduled to begin construction within 2025. Three plants are planned for construction in Arizona, U.S.A., with the first plant starting mass production in the fourth quarter of 2024. Additionally, TSMC announced a $100 billion investment in the U.S. for March 2025. In August 2024, construction began on TSMC’s first plant in Europe, located in Dresden, Germany. However, even once those factories are operational, the majority of advanced process manufacturing will still take place in Taiwan.
I was genuinely surprised, as I had never imagined that TSMC would build a factory in Japan. I think this decision would not have been possible without government support measures such as subsidies, and important customers. I also believe that the change in global trends, as described above—with a close focus on the concept of economic security—was a significant reason. I think government support during this time would not have been possible without a change in global trends. Many people in the semiconductor industry may not have thought that TSMC would really build a factory in Japan.
Construction site of Japan Advanced Semiconductor Manufacturing (JASM, TSMC Japan Fab 23), TSMC’s first plant in Japan
(Left: December 2022; Right: December 2023, Kumamoto)
Note: According to TSMC’s website as of November 1, 2023, TSMC’s total wafer manufacturing capacity in 2022 was 15 million 12-inch-equivalent wafers per year, including subsidiaries. This translates to a monthly production capacity of 1.25 million wafers. The production capacity of Japan Advanced Semiconductor Manufacturing (JASM, TSMC’s Kumamoto plant) was 55,000 wafers per month, accounting for about 4% of TSMC’s total capacity. While this may not seem significant, considering Japan’s 6% (monetary) share of TSMC’s revenue in Q3 2023, it represents a relatively significant scale for Japan.
It is sometimes argued that Japan’s semiconductor industry declined because it failed to adapt to horizontal specialization. But is that really the case? As mentioned above, Samsung and SK Hynix, South Korea’s two largest semiconductor companies, are both IDMs. The South Korean semiconductor industry’s success is not due to its response to horizontal specialization. To begin with, the top two in the semiconductor revenue ranking of 2022 are IDMs, and half of the top ten are IDMs. Japanese companies dominated the DRAM industry in the 1980s, but lost out to two South Korean companies and Micron in the U.S. This is unrelated to horizontal specialization. Similarly, Intel and Texas Instruments (TI), two U.S. IDMs, failed in the DRAM market and withdrew from it. However, they switched to other products and are still thriving as IDMs as of 2025.
As mentioned in Volume 15, the foundry model originally emerged from Taiwan’s national policies and its circumstances at the time. So, it’s understandable that it did not originate in Japan.
Moreover, it is not true that Japanese companies do not make use of horizontal specialization. Renesas, a leading Japanese IDM, has also use foundries for its fine processes since a certain period. So, it is clear that they are indeed making use of horizontal specialization. Many other Japanese IDMs, including ours, use foundries and OSATs. Since around 2000, our company has been using overseas foundries and OSATs. Japan also has fabless semiconductor companies, which similarly make use of foundries and OSATs. A foundry and OSAT can be used to manufacture products if an idea is available. The same is true for companies everywhere.
As I discussed in Volume 14, the fabless companies that started up in the U.S. in the 1980s simply became independent or started their own businesses to realize their ideas. This does not mean that semiconductor companies were not split into design companies (fabless companies) and manufacturing companies (foundries). As far as I know, AMD is the only IDM in the top revenue ranking that has actually separated its manufacturing division and made that division a fabless company.
Taking the above into consideration, it’s just my personal opinion, but I still feel it’s not quite right to say Japan has fallen because it could not cope with horizontal specialization.
This concludes my discussion of major trends in semiconductors. I did not expect to write eighteen articles on miniaturization, shift to larger diameter silicon wafers, and horizontal specialization. The more I looked into it, the more it expanded, and here I am. Thank you for your attention. I will continue to cover semiconductor-related topics, so please stay with me.
In the next post, I plan to talk a little more about “advanced logic semiconductors,” which I introduced in the previous volume as the product where the horizontal specialization is advancing.
Click below to read this series.
Semiconductor Miniaturization:
Volume 1: Semiconductor Miniaturization: What is Moore’s Law?
Volume 2: Semiconductor Miniaturization and Manufacturing Process
Volume 3: Semiconductor Miniaturization and International Technology Roadmap
Volume 4: Semiconductor Miniaturization and Semiconductor Business
Volume 5: Semiconductor Miniaturization and Semiconductor Business (Part 2)
Volume 6: Semiconductor Miniaturization and Semiconductor Devices
Volume 7: Semiconductor Miniaturization: What is MOSFET Scaling?
Volume 8: Semiconductor Miniaturization: Limitations of MOSFET Scaling
Volume 9: Semiconductor Miniaturization and Analog Circuits
Shift to Larger Diameter Silicon Wafers:
Volume 10: Shift to Larger Diameter Silicon Wafers: How a Common Material, Silicon, Became a Main Player
Volume 11: Shift to Larger Diameter Silicon Wafers (Part 2): How Silicon Wafers Are Made
Volume 12: Shift to Larger Diameter Silicon Wafers (Part 3): Reasons and History
Horizontal Specialization in the Semiconductor Industry
Volume 13: Horizontal Specialization in the Semiconductor Industry and the Rise of Fabless Companies
Volume 14: History of Horizontal Specialization in the Semiconductor Industry–Emergence of Fabless Semiconductor Companies
Volume 15: History of Horizontal Specialization in the Semiconductor Industry–Emergence of Foundries
Volume 16: Horizontal Specialization in the Semiconductor Industry—Is a Foundry Just a Subcontractor?
Volume 17: Horizontal Specialization in the Semiconductor Industry—Horizontal Specialization by Product and Country (Part 1)
Volume 18: Horizontal Specialization in the Semiconductor Industry—Horizontal Specialization by Product and Country (Part 2: By Country)