|Date||January 27, 2023|
|Speaker||Chris MILLER (Associate Professor, International History, Fletcher School, Tufts University)|
|Commentator||WAKABAYASHI Hideki ((Professor, Graduate School of Management, Department of Management of Technology, Tokyo University of Science)|
|Commentator||OTA Yasu (Columnist, NIKKEI)|
|Moderator||NISHIKAWA Kazumi (Director, Policy Planning and Coordination Division, Commerce and Information Policy Bureau, METI)|
Semiconductor technology, which has shaped modern society and infrastructure, is linked with advances in military applications. The development, procurement and supply of semiconductor technology has become a critical priority for policy makers worldwide, who recognize its fundamental role within current society and the future development and training of Artificial Intelligence for civilian and military use. Professor Chris Miller, author of “Chip War” looks at lessons from the Russian war in Ukraine, and the future of computerized warfare, which is especially relevant in the face of Chinese technological and military advances that could threaten the stability of vital supply chains in the Taiwan Straights. Wider geopolitical tensions are reflected in increased diversification and politicization of semiconductor supply chains. This, along with ongoing shifts in the broader electronics industry will have wider economic and political implications, both in Japan and worldwide.
The History of Semiconductors
The semiconductor chips we associate with consumer devices like smartphones and PCs, actually emerged from a United States commitment to miniaturizing computing power during the Cold War. The first semiconductors were applied to military systems, and a deep relationship between defense industries and advances in semiconductor technology has continued. It was only in the 1970s onwards that consumer electronics and corporate computers started consuming a larger share of chips, although defense industries still played a large role in buying chips and R&D. In 1965, Gordon Moore, co-founder of Intel, noted that the number of transistors that fit on a chip double every year or two. This observation, known as Moore's Law, has continued at a similar rate, enabling the miniaturization of computing we know today. This is relevant for both civilian and military systems, that demand ever smaller transistors and ever more advanced semiconductors.
In the 1970s, the first precision guided strikes took place during the Vietnam War, revolutionizing military attitudes to technology. Over the past 50 years, almost every military system has started to include an extraordinary amount of computing power in terms of processor and memory chips required for sensor data processing. The Russia-Ukraine war has highlighted that communications sensors and precision guided munitions continue to transform the way wars are fought, and many key tactics that have worked for Ukraine in this war stem from capabilities produced by semiconductors. For example, even older systems like the Javelin anti-tank missile from the 1970s and 1980s, have several hundred semiconductors. and Ukraine leans very heavily on high-tech military weapons, communications, and cyber defense systems that are enabled by advances in computing power. For Example, Starlink, the low earth orbit satellite system started by Elon Musk, gave Ukrainians access to widespread satellite communications that were unable to be cut off, jammed or hacked by Russia. Militaries today demand long-distance, secure communications systems enabled by semiconductors, such as HIMAR precision guided missiles. The US and others give Ukraine an extraordinary amount of signals intelligence, processed into coordinates of potential Russian targets. When entered into the HIMAR guidance computer, the missile does the rest of the work. This computerized warfare wouldn't be possible without semiconductors. Advanced computing capabilities are also at the core of Ukraine's cyber defenses. Government data is stored on cloud computing systems defended by Microsoft and Google, which allowed it to stay functional in the early days of the war. These cloud data centers, vast warehouses full of semiconductors, are protected by some of the most advanced cyber defense systems ever produced.
Defense planners and policymakers today in Washington and Beijing envision battlefields of the future to depend even more on sensors, communications capabilities, and computing power. Militaries transfer and process huge quantities of data, from rapidly increasing numbers of infrared, LIDAR, radar, and optical sensors. Some military systems are gaining the partial autonomy to fly themselves, or automatically locate and identify potential targets. Such advancements require new semiconductors, both for the autonomous systems, and for the GPU chips used in the data centers to train these autonomous systems. With the US and China both considering this, there is an inevitable relationship between computing technologies, intelligence capabilities and military power, reflected in accelerating geopolitical tensions between China and its neighbors. A number of countries have imposed restrictions on investment, transfer of information or workers, and the export of certain types of chips and machines into China. Meanwhile China is spending very heavily to boost its chip industry and make it more self-sufficient. Most advanced semiconductors require machine tools, materials, designs and software from the Netherlands, Japan, and the US, with Taiwanese or Korean fabrication capacity. China plays a role in lagging edge, less sophisticated semiconductor orders, relying on imported machine tools. Since 2014, semiconductors have been a core priority for China's economic policy, leading to increased capacity to produce less sophisticated chips, but its dependance on imports of advanced chips and machine tools, chemicals, designs, and chip design software continues. The rest of the world is also reliant on a few facilities to produce semiconductors. Around 90% of advanced processor chips for smartphones, PCs, data centers and telecoms infrastructure are produced in Taiwan by TSMC (the Taiwan Semiconductor Manufacturing Company) with the other 10% produced by South Korea’s Samsung.
Supply Chain Resilience
As tensions rise between China and Taiwan, the concentration of cutting edge chipmaking in Taiwan is a growing source of concern. The US is limiting the ability of Chinese companies to acquire US tools or software, and other international chip firms are also changing investment patterns. Semiconductor manufacturing countries integrate themselves very deeply into international supply chains, but this is an increasingly difficult strategy for Chinese firms. Therefore, the government is implementing massive capital investment into national and local subsidies. It is questionable whether China can sustain this level of funding, and any domestic cutting-edge supply chain would likely spill over into the military sphere as tensions escalate. With Chinese pressure and military power intensifying each year, the risk of a blockade or conflict in the Taiwan Straits that could impact Taiwan's ability to export semiconductors is growing. This has sparked debate about what governments can do to provide resiliency in the case of a Taiwan crisis that would disrupt semiconductor supply chains with costly effects worldwide. Governments and companies are rethinking their reliance on a small number of geographically concentrated chip makers, as reflected in policy and corporate investment trends. The US has passed legislation such as the Chips and Sciences Act, devoting $40 billion to manufacturing and over $10 billion to R&D, to reduce the cost gap between producing in the US versus Taiwan or South Korea. Industrial cost differentials between facilities are driven by tax policy differentials, as costs for specialized machine tools and labor do not differ significantly. The CHIPS Act is intended to provide a generous tax policy to make it competitive to produce semiconductors in the United States, with the goal of supply chain resilience. DARPA, the Defense Department's R&D arm, has launched an electronics resurgence initiative to fund cutting edge chip production, and new controls have been imposed on transfer of technology to China. These measures are intended to consolidate semiconductor technology leadership in the United States and allied countries and make supply chains more resilient in case of a crisis. Other major governments are also diversifying the geography of semiconductor fabrication. India is subsidizing lagging edge fabrication, the European Union is preparing a large Chips Act, Japan has promoted the opening of a new facility by TSMC, and Taiwan and South Korea are providing tax credits. Companies in the US, Japan and Europe are beginning to change their investment patterns, and TSMC is as a result of this is building new facilities in the US, Japan and possibly Germany.
Customers are also demanding supply chain resilience. Dell is phasing out Made in China chips by 2024. Apple is expanding assembly capabilities in Vietnam and India. Mexico is trying to attract some assembly capacity for computers, and India is building a semiconductor fab. Across the electronics industry, there is more focus than ever on supply chain diversification. The chip industry itself is also changing. Moore's Law used to cheaply provide doubled computing power each year, but this is no longer the case. Advanced semiconductor production costs have not declined, so designing chips is becoming more expensive for companies. This could concentrate the industry and deter innovation, so open-source architectures are trying to provide cost-effective production methods capable of comparable advances. Computer architectures have also shifted, and specialized chip designs like GPUs optimized for artificial intelligence offer an opportunity for an industry reset and for new companies to rise based on their capabilities. Packaging of multiple different chips together to provide faster interconnect speeds with better performance is another key trend in the chip industry, making this traditionally low value, less capital-intensive packaging side of the industry, increasingly important.
A driver of growth for the future of the semiconductor industry is in data centers and cloud computing capabilities, which we increasingly rely on in our daily lives. Automobiles are another major growth segment, as electric vehicles gain popularity, with autonomous driving features requiring advanced semiconductors. The Moore’s Law Era of general-purpose chips is becoming a differentiated landscape, with chip design, software, tools and packaging arranged in different ways. This means traditional players are finding it harder to compete. Tech firms in the semiconductor space and their customers in the electronics industry must adapt to technological shifts while simultaneously de-risking supply chains from excessive dependance on China and Taiwan.
As an analyst in Japan in the nineties, I felt a sense of urgency towards Japanese semiconductor companies. I was troubled by the friction between Japan and the United States. Chip War answered some of my long-term questions, and considers the semiconductor industry from the technological perspective, and also from wide ranging management strategy, finance, education, supply chain, and national security perspectives. Japan should pay attention to the perspective of fairness and contribute to the world economy and to western countries. My one correction is that NAND flash was invented by Dr. Masuoka, a former professor at Tohoku University. Also, it was Toshiba, not Intel, that put it into practical use. Other than that, this book was very informative. The world industrial structure consists of layers of finance, software, science and technology, and manufacturing. I feel that Japan could contribute to the world on a production basis, and I am hopeful for a Japanese semiconductor comeback, that would contribute not only to B2C (Buyer to Consumer) markets, but to B2G (Business to Government) and B2B (Business to Business) markets.
I have a couple of observations as a journalist. The West is a fuzzy notion. Can the US, EU and Japan include Taiwan in “The West” when Taiwanese people have relatives and families in mainland China, with significant investment, communication and movement of people between them? We cannot really assume that Taiwan is included. However, the founder of TSMC, Morris Chan, is a strong critic of the CCP, although he also criticizes the US, saying that free trade has come to an end because of their actions. People in Taiwan don't want to be seen as a 51st state of the United States. and the incumbent DPP (Democratic Progressive Party) lost the November local elections. What about Singapore, Malaysia, Vietnam or even Indonesia? There is a dense supply chain crisscrossing East Asia, including ASEAN member states, leading to the question of alliances. The West doesn't appear strong enough to create a safe supply chain, so there needs to be more convincing narratives and tangible economic benefits. I would also like to talk about economic stability. Industries need to spend a lot of money, and who will pay these costs? Consumers are not ready for a cost-push semiconductor inflation, so we will see significant amounts of money poured into unsustainable government subsidization. Japan, along with the US and China, is already in a deep fiscal deficit and cannot keep spending money on industry forever. Each nation in the complex Asian supply chain must consider economic sustainability, or we will all fall together. While the military threat is imminent, economic risk is also imminent. We need to draw a clear line defining sensitive technology versus regular technology for which free trade is intact.
I did some research into Russian semiconductors. There are two or three major companies, that are very good at semiconductor design due to a post-cold-war system that is independent from western technology. While they have design capability, they needed to ask TSMC to produce the chips. However, with the supply chain cut between Russia and Asia, it seems they are now importing from China. I checked several years of trade between Russia and China, and it seems China has started exporting a significant number of chips. This is not quite the situation we expected, and it puts China in a better position politically in Russia, representing a change in the structure of their relationship.
Prof. Wakabayashi spoke about Japan’s individual respective role in global collaboration, and Mr. Ota highlighted the difficulties of alliances and economic stability. Do you have any responses to these comments?
Regarding the international supply chain, no country can be self-sufficient when it comes to semiconductors. Making advanced chips requires intellectual property, software, and machine tools from multiple countries. International supply chains have been key drivers of technological advance by allowing investment and specialization. Certain countries have specialized capabilities, the United States in chip design, Europe and Japan in machine tools and materials, Taiwan and Korea in fabrication. This has worked well but created risk of overconcentration in specific geographies. In terms of “alliances” governments and companies have complex views. However, in Europe, the US and Japan, there has been great concern about supply chain risks in the Taiwan straits in the sense that something must be done to reduce our reliance, and each of these countries are making efforts towards this. Of course, managing relations with Taiwan is complex, and countries are trying to become less dependent on Taiwanese semiconductors while simultaneously defending Taiwanese autonomy from Beijing. Economic risks are involved, with government spending in the chip industry, potentially risking overcapacity in the future, and while there will always be demand at the leading edge, there is scope for overcapacity in memory and lagging edge logic. I think a big risk is the lagging edge logic build-out underway in China right now, as there has been little consideration of commercial viability or profitability, and sizeable government subsidies. On the differentiation of civilian and military technology, in the past we knew what military technology was. With artificial intelligence, the same chips that train a car to drive autonomously can also train a drone to fly. How do we regulate this, when there is no differentiation of what is civilian and what is military? That is a key question facing us right now, and this dual use is a key challenge for export controls that previous generations of regulations didn't need to address.
When thinking about semiconductors, geopolitical outlook is very important. How long do you think the U.S.- China Semiconductor conflict will last? How will it end?
The key driver of the semiconductor conflict is the military balance in East Asia. This military balance has swung from the US to China, which has built more advanced military systems and expanded its navy, air force, and missile forces. As China's military has expanded quantitatively, it's become more important for the US to maintain its edge qualitatively. That's why the US establishment is set on retaining advantages in semiconductor capabilities, to achieve military systems that are good enough to offset the sheer number of Chinese military systems in the Taiwan Straits. Until the military situation becomes clearer, technological tensions are unlikely to diminish, as both key parties are trying to deploy advanced technologies to military systems.
Especially cutting-edge semiconductors have military uses, but also civilian applications. Is there a middle ground for restrictions that would prevent China from military use of advanced chips while not hampering their innovation?
The US Government would give the following answer. There's only one type of chip that is prohibited for export to China, which is the GPU chips that are used to train AI systems. Chinese private sector firms like Alibaba or Tencent are not prohibited from taking their data to a data center in Singapore, India, Japan or anywhere else and running it using the most advanced chips. Current restrictions are only on the transfer of chips to China. That's a middle ground of a sort, but China will most likely decide not to allow its firms to move data abroad for processing. However, I don't know of a more viable middle ground, because if these advanced GPU chips are transferred to China, there's no way to ensure they are used for civilian purposes.
There are some other questions focusing on US trust in Japan as well as Korea or Taiwan. What kind of role does the U.S. expect from them in future?
There is already specialization in the roles that Korea, Taiwan and Japan play in the chip industry. Korea specializes in memory chips, Taiwan, in processor chips, and Japan in unique capabilities especially in certain materials and machine tools. However, in a regulatory or a political sense, they fall into different categories. I would note that Japan and Korea have a security alliance with the United States. Taiwan does not, which is an important factor in in this dynamic. Views on Taiwan risk and China risk differ in Japan, Korea and certainly in Taiwan. The concept of Taiwan being a risky place to produce large numbers of advanced semiconductors is obviously something that its government and populace disagree with. Despite this difference in views, when thinking about how to stabilize the supply chain, South Korea and Japan share some agreement that advanced logic capacity is too concentrated in Taiwan. In this sense, there is some alignment between them and the United States on the nature of the challenge.
“Chip war” discusses semiconductor business strategies such as Tick-Tock strategy, copied technology and chip-first technology. For the US, its biggest market is also its biggest competitor, and so business strategy needs to change viewpoints, concentrating not only on profits but also on public life, supply chains, and global technology contributions. In the future I would like to hear your perspective on the changes that are occurring in semiconductor business strategies, where core business factors must be considered.
Technological funding must be discussed at some point because the current level is simply not sustainable. A line needs to be drawn between what we can protect and what we don't need to protect, and this is my big question for policymakers and political leaders in the West.
There are many geopolitical and economic issues to be considered, as well as various economic and security considerations regarding commercial and military use. The government should think on this from long-term perspective with our global friends.
*This summary was compiled by RIETI Editorial staff.