Mathieu Duchâtel
There are striking similarities between Europe’s and Japan’s semiconductor industries, and between how the European Commission and the Japanese government are reinvesting in their industrial policy to cope with an international environment characterized by increased US-China rivalry.
Last June, Japan’s Ministry of Economy, Trade and Industry (METI) released its Strategy for Semiconductors and the Digital Industry. According to METI, semiconductors are for the digital revolution not what oil was in the 1970s or the steam engine during the industrial revolution - instead they are the "rice of the industry, essential and irreplaceable to all industries." The metaphor combines an evocation of national survival - everyone needs rice to survive - and a vision of Japanese pride, as one of the world’s producers of rice of the highest international standard.
Japan’s industrial survival at stake
Like Europe, Japan has serious vulnerabilities but also key strengths, including segments of global leadership. This survival notion reflects Japan’s industrial decline in semiconductors, as compared to the United States and to manufacturing powerhouses in Taiwan and South Korea. According to METI, Japan’s share of the world’s market shrank from 50% in 1990 to 10% today. During the same three decades, Europe’s share of semiconductor manufacturing capacity also experienced a similar stratospheric decline, from 44% to 9%.
METI’s stated goal is to maintain that remaining 10% market share by 2030. This will require considerable investment - up to JPY 5 trillion (EUR 38 billion) according to the government’s strategy paper. That is less than the USD 52 billion government funding currently being discussed in the United States. It is arguably also less than the combination of the EU policy instruments put in place to support the European semiconductor industry: the EU allocated 20% of its EUR 750 billion recovery package to digital transformation (though only a portion of that will fund projects by European semiconductor firms). The EU is also setting up a public-private semiconductor alliance to secure additional investment and is preparing a second Important Project of Common European Interest for microelectronics. But METI’s number is an estimate, not an approved budget. Japan will adjust to operate on a scale comparable to Europe and the United States.
METI’s analysis of the reasons behind Japan’s decline is clear-eyed and direct. It was to some extent forced upon Japan by the United States during the 1986 trade negotiations: the US-Japan Semiconductor Agreement imposed anti-dumping guarantees and a 20% market share provision opening the Japanese market to foreign producers over a five-year time frame.
More decisively however, the Japanese industry missed two transformations in the late 1980s. It failed to invest in logic chips and instead chose to focus on DRAM memory, at a time when the US industry was precisely giving up memory to focus on computing power. As a result, Japanese companies were exposed to fierce Korean competition in memory chips. And, just like Europe, Japan missed the train of the horizontal division of labor between Integrated Circuit (IC) design and IC manufacturing, a transformation that originated in the Silicon Valley and in Taiwan. California saw the rise of fabless giants. Broadcom, Qualcomm and NVIDIA captured the upstream of the global value chain with their design of new processors for consumer devices. In Taiwan, the Taiwan Semiconductor Manufacturing Company (TSMC) created the contract foundry model in 1987, patiently building a position of global leadership based on constant innovation in manufacturing capacities. Japan, on the other hand, failed to keep hold of its giants or create new ones.
Japan’s supply chain risks
The legacy of the late 1980s was already a challenging one to overturn, but Japan, like Europe, could take comfort in the fact that a global value chain was organized on the principle of rationality of market actors, with interdependence ensuring access. This is no longer the case. Covid-19 caused major disruptions in the global semiconductor supply chain and the US-China tech war has also changed Japan’s threat assessment.
The current shortages have been a wake-up call. The global car industry may lose up to USD 110 billion in 2021 due to the current shortage of supplies. This affects Japan’s giant automotive industry, even if it was better prepared than European carmakers thanks to the resilience measures put in place after the 2011 Tohoku earthquake and tsunami. Toyota, for example, fully benefits from its chip stockpiling policy. Today, shortages have expanded to a larger range of products in the digital economy, from smartphones to servers and telecommunication hardware. For Japan, geopolitical competition, rather than natural disasters or global health crises, is the new risk to address in order to avoid losing access to items that are essential to the country’s economy.
For Japan, geopolitical competition, rather than natural disasters or global health crises, is the new risk to address in order to avoid losing access to items that are essential to the country’s economy.
This perspective gives way to a vision which stands in contrast to the EU’s emphasis on strategic autonomy. METI’s starting point is Japan’s relative power in the international system, and its need to resist geopolitical shockwaves. To answer this need, Japan’s policies should position the country as an indispensable actor in the global supply chain: Japan itself should own chokepoints to immunize itself from hostile leverage. In this context, Japan wants to be an indispensable partner in a triangular interdependent relationship with the United States and Europe. This resolutely Western outlook is a strategic choice.
The other three Asian semiconductor players matter for other reasons. Taiwan matters as a key partner. Korea is mainly pictured as a competitor in the memory segment, while China’s state capitalism and hostile intentions pose a quasi-existential threat.
There is a certain irony in how, through a twist of history, the risk of Chinese dumping intentionally bankrupting segments of Japan’s semiconductor industry is now seriously considered by a country that 35 years ago was the object of US trade sanctions for its dumping practices. Times have completely changed. The Chinese Communist Party has formulated a vision of global leadership, even though the Chinese semiconductor industry is extremely dependent on foreign imports. As China inevitably catches up, it will acquire huge strategic leverage if it owns chokepoints in the semiconductor industry. Japan successfully managed China’s 2010 cutoff of rare earth exports, but semiconductors could cause more damage.
This risk assessment leads to a logical conclusion heard also in Europe: what matters most is owning chokepoints and cultivating strengths. By definition, Japan’s Integrated Device Manufacturers (IDMs, or companies that design and produce semiconductors), just like European firms, do not aim at mastering the cutting edge nodes of semiconductor manufacturing. The stories of the semiconductor branches of Toshiba, Fujitsu, Hitachi and Mitsubishi have similarities to those of Europe’s three largest IDMs: STMicroelectronics, Infineon and NXP. Not being TSMC does not mean that they have not built thriving businesses. They still have world leading strengths in microcontrollers, sensors and power semiconductors. METI recommends investment in R&D and capital expenditure to maintain Japan’s highly competitive position in power semiconductor and sensors. For microcontrollers for the automotive industry and smart factories, METI calls for supply chain diversification in order to maintain Japan’s high market share. And Japan holds a strong card with its 50% market share in many categories of materials for semiconductors. In 2019, it showed that it was ready to play this card when it imposed restrictions on exports of photoresists to Korea.
The advanced foundry debate in Japan and Europe
METI’s strategy paper underlines the importance of enabling domestic production for advanced logic semiconductors. This foundry issue will be very familiar to European observers who, since last year, have experienced a fierce debate in Europe between those who want an advanced foundry on European soil, crossing the threshold of the 7 nanometer manufacturing process, and those who do not.
Europe’s most advanced fab, in Dresden, uses 22 nanometer process technology. Japan’s most advanced foundry process is 40 nm. Soon, the cutting edge will be 2 nanometers - the Taiwanese government has just greenlighted the construction of TSMC’s new fab in Hsinchu. Because the market for the most advanced logic processors is essentially state-of-the art consumer digital devices, such as the iPhone, many industrial players in Europe argue that subsidizing an advanced fab would be a waste of public money. Their argument is well captured by the head of Global Foundries’s most advanced fab in Dresden, when he states: "We believe that a Dresden technology corridor from 55 to 22/12 nanometers will provide solutions for 90 percent of the chips that European industry will need well after 2035. We have the solutions ready or in the pipeline that will be needed in the next 15 years".
The question has two separate dimensions. One is to maintain stable access for the European and the Japanese automotive and other innovative industrial consumers of semiconductors, when demand spikes or when the supply side is disrupted overseas. For that, crossing the 7 nm threshold is not needed, but having more advanced local fabs would ease the pressure on the designers as they seek greater performance for their products.
Second, there is the issue of what the industrial landscape will look like in twenty years. From that perspective, the question of advanced manufacturing can’t be brushed away. In five years, the production of logic chips using 7 nm processes and below will be concentrated in Korea, the US and Taiwan. China will still face technological bottlenecks due to restrictions on technology transfers, but it will use all its power to overcome such hurdles. There will be more IC design companies in Japan and Europe in need of advanced process manufacturing.
China will still face technological bottlenecks due to restrictions on technology transfers, but it will use all its power to overcome such hurdles.
They could purchase chips from TSMC, including possibly from its new 5 nm fab in Arizona. Samsung and Intel may also offer options. But the risks of disruption exist, especially in a situation where the global demand could exceed the capacities of the fabs.
This presents Europe and Japan with the same dilemma. Should they focus on their existing strengths and on owning chokepoints? Should they prioritize supporting the emergence of cutting-edge IC designers, to create demand on their soil for advanced manufacturing processes? Should they subsidize TSMC or Intel to build an advanced fab under the assumption that a global market exists for such products, even if the Japanese and the European markets are currently too small to absorb production?
The answer to these questions will determine the future position of Europe and Japan in the global semiconductor supply chain. Both intend to double down on support for R&D and innovation for their SME and start-up ecosystem. The METI paper emphasizes "more-than-Moore" research, with the goal of creating circuits that combine logic, memory and sensor functions in a single chip. The Moore Law states that the number of transistors in an integrated circuit doubles about every two years, and empirically describes and predicts the miniaturization of industrial processes. But 2 nm (the distance between two transistors on a chip) might be the physical limit of the law, and of the new frontiers is vertical integration of layers on one chip. More-than-Moore has also long been a priority for European companies and research centers.
But there is a key difference in how the EU and Japan currently handle the matter. Japan has set the course for cultivating a long-term relationship with TSMC, starting relatively small, while the EU is pondering options and could still prioritize advanced logic IC design rather than manufacturing as a recipient of public funding.
Earlier this year, TSMC announced an investment in a subsidized R&D center in Tsukuba, near Tokyo. The facility brings together about 20 Japanese companies and Japan’s leading R&D public agency, the National Institute of Advanced Industrial Science and Technology (AIST), to work on vertical integration in advanced manufacturing. And it now appears that almost all lights are green regarding the construction by Taiwan Semiconductor Manufacturing Co. of its first foundry in Japan. TSMC’s Chairman Mark Liu has confirmed that the company is currently in "the due diligence process to have a specialty technology fab in Japan". Nikkei Asia has recently reported that the fab should become operational in 2023, using 28-nanometer technology mainly to produce image sensors for Sony. This is not advanced logic, but this is one more step towards rebuilding the Japanese semiconductor ecosystem with TSMC.
TSMC’s discussions with Japan are more advanced than the "very early stages" of the Taiwanese company’s exchanges with Germany. TSMC may in the end invest in Dresden to respond to the needs of the German industry. But what the Japanese approach illustrates is that supporting an industry’s existing strengths does not necessarily exclude the pursuit of large-scale industrial projects.
No comments:
Post a Comment