[East Asia Institute-Choi Soon-ho Academic Council-Seoul National University Institute for National Future Strategy Jointly Hosted Academic Forum] Changes in the International Order and Economic Security Strategies

YouTube Link: https://www.youtube.com/watch?v=JRM8fL3JjIs

Video Script

US-China Competition for Semiconductor and AI Supremacy and Korea's Economic Security

Yes, good afternoon. I am Seokjun Kwon from the Department of Chemical Engineering at Sungkyunkwan University, who was just introduced. I believe I am the only one from an engineering background here. In this third session, I will discuss the competitive relationships between countries in advanced industries, particularly focusing on South Korea, the United States, and China. Although I am not an expert in international politics, as you know, the logic of geopolitics is evolving, with the focus shifting to semiconductors and, more recently, AI. The current administration places a strong emphasis on AI sovereignty as a core national agenda. However, considering the context in which this strategy can open new breakthroughs for our industry, and the fact that not only our country but also the US, China, and other nations worldwide approach AI with a profound sense of supremacy consciousness,

I will begin my presentation by emphasizing the importance of examining these issues not only domestically but also within an international political context, particularly from an economic security perspective. Did it move on? As you are well aware, China is currently the biggest threat to South Korea's semiconductor industry. While China focused primarily on low-value-added semiconductor production until a few years ago, it is now showing signs of a full-fledged transition towards high-quality production.

Specifically, in the realm of AI semiconductors, China is nearing a stage where the internalization of semiconductor supply is largely achieved domestically. This has a profound impact on South Korea's semiconductor industry. It implies that the technological sanctions that the US has imposed on China, particularly the pinpoint sanctions aimed at preventing the production of core semiconductor chips, have largely been ineffective, or that China is developing strategies to circumvent or overcome them.

In China, the semiconductor industry's development appears to be synchronized with the periods of Xi Jinping's first, second, and third terms, with updates occurring approximately every five years. From the first Semiconductor Big Fund in 2014 to the second in 2019, and the recent third Semiconductor Big Fund in 2024, not only has the scale of the Big Fund itself increased, but the third phase, in particular, shows a significant application of semiconductor technology not only in the vast semiconductor industry but also in its practical applications, extending beyond AI semiconductors to mobility, power, telecommunications, and even energy and biotechnology.

One of the key emphases this year is the production of new quality, which ultimately signifies China's strong determination to move towards high-quality, high-value-added semiconductors. In this context, there are several key companies that form China's AI and semiconductor ecosystem. Besides the well-known Huawei, there are companies like DeepRoute.ai, which recently shocked many countries including South Korea, and Alibaba, which is developing powerful LLMs comparable to OpenAI and Claude. From South Korea's perspective, it is crucial to note that China is investing heavily in the semiconductor industry, particularly in semiconductor manufacturing. According to recent reports, foundries, which play a critical role in system semiconductor production,

are an area where China has secured more than three foundry companies that could rank within the top ten globally. What is even more interesting is that the influence of Taiwan, which has always been mentioned in the context of foundries, is gradually diminishing. While the position of dominant players like TSMC remains largely unshaken, the influence of supporting companies like UMC, PSMC, and Vanguard is declining, and Chinese companies are filling that void.

The reason I emphasize this is that while China's foundries may not yet possess the same technological capabilities as TSMC or Samsung Foundry, these foundry companies are rapidly expanding their capacity. Furthermore, their technological capabilities are also advancing. Consequently, in areas that do not require specific high-performance semiconductors like AI semiconductors or application processors, but rather in areas employing relatively mature legacy processes for industrial semiconductors, bio-semiconductors, general power, or communication semiconductors, the influence of Chinese foundries will grow. It is highly probable that they will capture over 10% of the global legacy foundry market within the next decade.

Moreover, what we must also note about these foundry companies is that their development does not stop at manufacturing alone. Foundries naturally require corresponding process equipment. Historically, China's weakness has been its high dependence on equipment from the US, Japan, and the Netherlands for most advanced processes. However, this dependence is also being addressed through triple subsidies for domestic equipment as these foundries expand in volume. These triple subsidies include, for example, government subsidies for foundry construction, subsidies for equipment manufacturers when foundries purchase Chinese equipment, and incentives for companies that purchase the chips. While the technological gap with the US remains significant,

we are observing a trend where this gap is rapidly narrowing. Another observation is the natural integration of the AI ecosystem based on semiconductors. When we discuss semiconductors and AI, we do not discuss them separately; we discuss them together. This is because no matter how well companies like Nvidia design GPUs, a corresponding manufacturing ecosystem is necessary for their physical production.

Without specialized foundries like TSMC, Nvidia cannot manufacture GPUs solely based on their designs. China is currently laying the groundwork for this and building a highway for a more internalized AI ecosystem. Taiwan plays a crucial role in this process. Not only does Taiwan possess foundries, but it also holds significant dominance in sub-10nm processes, which we commonly refer to as leading-edge processes.

This is why the concept of a Silicon Triangle, formed by Taiwan, the US, and China, with Taiwan at its center, has been prominent. The US aims to alleviate its reliance on Taiwan's advanced semiconductor production. One approach is to encourage major Taiwanese foundry companies employing leading-edge processes to relocate to the US as much as possible. For instance, TSMC is investing billions of dollars near Phoenix, Arizona, to build a fab utilizing 3nm process technology. Samsung Electronics is also constructing multiple fabs in Taylor, Texas, with a significant number of them likely to employ advanced processes below 5nm.

Nevertheless, in the Silicon Triangle formed by Taiwan, the US, and China, the extent to which China can catch up in advanced processes will be a key factor determining the economic security of this region, and by extension, the security of the Indo-Pacific. If the current trend continues, the US will secure at least 20-30% of its supply chain by the mid-2030s, a significant increase from the current less than 5%. This will reduce Taiwan's dominance. Another factor to consider is the extent to which China's sub-10nm fabs will increase during this period. Currently, China's share of sub-10nm foundry fabs is only about 2-3%. However, if current investments proceed as planned, it could reach at least 5% by the mid-2030s. Furthermore, Chinese foundries, particularly those linked to Huawei, may strengthen their dominance over Taiwanese foundries ranked 2nd, 3rd, and 4th, which may have weaker capital or competitiveness.

This is because Chinese foundries are rapidly increasing their capacity, and their technological capabilities are also improving. Therefore, in areas that do not require specific high-performance semiconductors like AI semiconductors or application processors, but rather in areas employing relatively mature legacy processes for industrial semiconductors, bio-semiconductors, general power, or communication semiconductors, the influence of Chinese foundries will grow. It is highly probable that they will capture over 10% of the global legacy foundry market within the next decade. Moreover, what we must also note about these foundry companies is that their development does not stop at manufacturing alone. Foundries naturally require corresponding process equipment. Historically, China's weakness has been its high dependence on equipment from the US, Japan, and the Netherlands for most advanced processes. However, this dependence is also being addressed through triple subsidies for domestic equipment as these foundries expand in volume. These triple subsidies include, for example, government subsidies for foundry construction, subsidies for equipment manufacturers when foundries purchase Chinese equipment, and incentives for companies that purchase the chips. While the technological gap with the US remains significant, we are observing a trend where this gap is rapidly narrowing. Another observation is the natural integration of the AI ecosystem based on semiconductors. When we discuss semiconductors and AI, we do not discuss them separately; we discuss them together. This is because no matter how well companies like Nvidia design GPUs, a corresponding manufacturing ecosystem is necessary for their physical production. Without specialized foundries like TSMC, Nvidia cannot manufacture GPUs solely based on their designs. China is currently laying the groundwork for this and building a highway for a more internalized AI ecosystem. Taiwan plays a crucial role in this process. Not only does Taiwan possess foundries, but it also holds significant dominance in sub-10nm processes, which we commonly refer to as leading-edge processes. This is why the concept of a Silicon Triangle, formed by Taiwan, the US, and China, with Taiwan at its center, has been prominent. The US aims to alleviate its reliance on Taiwan's advanced semiconductor production. One approach is to encourage major Taiwanese foundry companies employing leading-edge processes to relocate to the US as much as possible. For instance, TSMC is investing billions of dollars near Phoenix, Arizona, to build a fab utilizing 3nm process technology. Samsung Electronics is also constructing multiple fabs in Taylor, Texas, with a significant number of them likely to employ advanced processes below 5nm. Nevertheless, in the Silicon Triangle formed by Taiwan, the US, and China, the extent to which China can catch up in advanced processes will be a key factor determining the economic security of this region, and by extension, the security of the Indo-Pacific. If the current trend continues, the US will secure at least 20-30% of its supply chain by the mid-2030s, a significant increase from the current less than 5%. This will reduce Taiwan's dominance. Another factor to consider is the extent to which China's sub-10nm fabs will increase during this period. Currently, China's share of sub-10nm foundry fabs is only about 2-3%. However, if current investments proceed as planned, it could reach at least 5% by the mid-2030s. Furthermore, Chinese foundries, particularly those linked to Huawei, may strengthen their dominance over Taiwanese foundries ranked 2nd, 3rd, and 4th, which may have weaker capital or competitiveness.

US AI Supremacy Strategy and Korea's Role

These developments have a high probability of altering the geopolitical landscape of the Silicon Triangle. The US is concretizing numerous supremacy strategies for AI from a national policy perspective. The Stargate Project, announced immediately after the Trump administration began earlier this year, is an example of the US strategy. This Stargate Project alone amounts to $1 billion, with a significant portion of the fund being invested directly or indirectly by the US government, although private companies also participate. Just yesterday, the White House Office of Science and Technology Policy released a white paper titled 'Winning the Race: America's AI Action Plan,' which reveals a more explicit US AI supremacy strategy. I call it explicit because the US has already been implementing export control policies on key AI assets like GPUs, categorizing its allies and countries sharing core interests. According to the recently updated strategy, it is now highly likely that the US will compel key allies to participate in the AI technology ecosystem led by the US. Simultaneously, it intends to further strengthen control policies towards certain key competitors, which, as everyone knows, is China. Therefore, the accessibility to the AI ecosystem that the US speaks of ultimately implies convergence towards standards.

Naturally, when discussing standards, we must consider not only the AI model and ecosystem but also the next stage. The next stage here refers to the ripple effects of AI's application across various domains such as manufacturing and defense industries, not just AI itself. Therefore, although the US states that direct production is currently difficult and reshoring is challenging, it is believed that they ultimately view this as a game-changer. Notably, the US states in that white paper that AI regulation and safety will be temporarily set aside. What this means is that it is now more important to actively utilize AI, and to this end, they intend to remove most government and private regulations and maximize performance, effectively paving the way for rapid advancement.

Maintaining these AI supremacy strategies requires a very strong semiconductor foundation. Therefore, if South Korea intends to strengthen cooperation with the US from an economic security perspective, I believe South Korea should propose the following to the US. For example, the US cannot yet fully produce advanced semiconductors independently, making Taiwan the country with the greatest reliance in East Asia. However, Taiwan is not a US ally and has no diplomatic relations. Therefore, South Korea and Japan, which can form an alliance with the US and play a key role in the supply chain—Japan's semiconductor manufacturing base has weakened considerably, making South Korea, from a relative and technological perspective, the most suitable partner for the US's AI supremacy strategy.

can become the most crucial technological partner. This is because the US cannot form a partnership with China. These are points we can consider. So, what is China thinking? We need to examine this. As I have already mentioned extensively, China has only one objective: to persevere until the end. Persevere in what? In internalizing the technologies that the US is sanctioning in semiconductors and AI, one by one. Naturally, technological gaps still exist during the internalization process. Broadly speaking, they have only reached about 70%, but as seen in the DeepRoute.ai shock in January of this year, if overcoming US sanctions is impossible, China will find a way around them. If circumvention is impossible, they will find a way to break through. They conduct various experiments in this manner. They have sufficient capital for such experiments, and more importantly, an overwhelming number of specialized personnel to carry them out. When ranking the top 100 AI research institutions globally, approximately 70% are Chinese institutions or Chinese individuals. In other words, global AI technology has reached a point where it cannot function without Chinese individuals.

China's AI Technology Internalization and Challenges

Ultimately, the most crucial technological partner. This is because the US cannot form a partnership with China. These are points we can consider. So, what is China thinking? We need to examine this. As I have already mentioned extensively, China has only one objective: to persevere until the end. Persevere in what? In internalizing the technologies that the US is sanctioning in semiconductors and AI, one by one. Naturally, technological gaps still exist during the internalization process. Broadly speaking, they have only reached about 70%, but as seen in the DeepRoute.ai shock in January of this year, if overcoming US sanctions is impossible, China will find a way around them. If circumvention is impossible, they will find a way to break through. They conduct various experiments in this manner. They have sufficient capital for such experiments, and more importantly, an overwhelming number of specialized personnel to carry them out. When ranking the top 100 AI research institutions globally, approximately 70% are Chinese institutions or Chinese individuals. In other words, global AI technology has reached a point where it cannot function without Chinese individuals. China is gaining confidence in these aspects, and particularly the private companies at the core of China's semiconductor AI ecosystem, such as Huawei, Alibaba, Baidu, Xiaomi, and SMIC, are rapidly internalizing most of the full-stack ecosystem for semiconductors and AI that we can imagine. And these internalized technologies are undergoing a quality transition, as mentioned earlier.

They find workarounds. If workarounds are impossible, they find ways to break through. They conduct various experiments in this manner. They have sufficient capital for such experiments, and more importantly, an overwhelming number of specialized personnel to carry them out. When ranking the top 100 AI research institutions globally, approximately 70% are Chinese institutions or Chinese individuals. In other words, global AI technology has reached a point where it cannot function without Chinese individuals. China is gaining confidence in these aspects, and particularly the private companies at the core of China's semiconductor AI ecosystem, such as Huawei, Alibaba, Tencent, Baidu, Xiaomi, and SMIC, are rapidly internalizing most of the full-stack ecosystem for semiconductors and AI that we can imagine. And these internalized technologies are undergoing a quality transition, as mentioned earlier.

They find workarounds. If workarounds are impossible, they find ways to break through. They conduct various experiments in this manner. They have sufficient capital for such experiments, and more importantly, an overwhelming number of specialized personnel to carry them out. When ranking the top 100 AI research institutions globally, approximately 70% are Chinese institutions or Chinese individuals. In other words, global AI technology has reached a point where it cannot function without Chinese individuals. China is gaining confidence in these aspects, and particularly the private companies at the core of China's semiconductor AI ecosystem, such as Huawei, Alibaba, Tencent, Baidu, Xiaomi, and SMIC, are rapidly internalizing most of the full-stack ecosystem for semiconductors and AI that we can imagine. And these internalized technologies are undergoing a quality transition, as mentioned earlier.

Building Korea's AI Ecosystem and Sovereign AI

is important. In particular, from South Korea's perspective, we need to consider how to approach AI. Until now, the general consensus has been that large-scale models are very important. However, going forward, the importance of AI in terms of what kind of reasoning is possible for whom, and what kind of applications are feasible, will increasingly grow. In particular, in this process, not only expensive GPUs for building large AI models but also purpose-built AI semiconductors specialized for tasks are becoming increasingly important.

Therefore, the current monopolistic structure of the global AI and semiconductor supply chain, dominated by Nvidia and further by a few companies like Nvidia, TSMC, SK Hynix, and Foxconn, is bound to be significantly alleviated. While this alleviation can be a crisis on one hand, it can also present a great opportunity for Korean companies. Particularly, the next battlefield for AI semiconductors is other manufacturing industries. This can extend beyond semiconductors to energy, biotechnology, shipbuilding, aviation, and steelmaking. South Korea, as one of the few developed nations capable of expanding its industrial base while controlling these impacts through democratic governance, can play a crucial role. To this end, South Korea is currently pursuing 'Sovereign AI'.

There are also discussions about building AI data centers in each region. The key challenge here is whether we can lead the infrastructure necessary to make such AI investments and establish the foundation. Power grids, industrial water supply, and communication networks are also very important, but these require long-term industrial policies rather than policies tied to a 5 or 10-year administration.

This will likely require significantly more funds and time than the construction of Incheon Airport or the KTX. Therefore, it needs to be pursued with a long-term perspective. Finally, one of the few things South Korea can do exceptionally well is to demonstrate to the world a case study where AI's role in manufacturing can generate new revenue. This can serve as a favorable bargaining chip in negotiations with the US. I apologize for taking up more time. I will conclude my presentation. Thank you.

AI Technology Innovation Ecosystem: State-Led vs. Emergent Challenges

Good afternoon. I am Park Jong-hyun from the Economic Security Cluster at the Institute for National Future Strategy, Seoul National University. Thank you for inviting me to this valuable event, and thank you for waiting until the end. Today, I will present on South Korea's economic security, particularly focusing on South Korea's economic transformation centered around AI and how to create a vision for South Korea amidst the US-China conflict. While various topics could be discussed, due to time constraints, I will focus on one key message through selection and concentration. For more detailed information, please refer to the book 'Economic Security Strategy for the New Government' published by our Economic Security Cluster, which has been sent to the printer. The summary of the current administration's AI pledge is 'Entering the Top 3 in Global AI'.

The US and China are likely to be the top two, so South Korea should aim to follow them. We need to expand private investment of over 100 trillion won plus government funding, and if necessary, establish a national fund. In terms of infrastructure, we need to purchase and pool 50,000 GPUs, establish data clusters in various regions, and secure power for these data clusters. We must secure sufficient AI social infrastructure capital. For talent development, establishing AI-focused colleges, reforming STEM education, and improving compensation are all very positive directions. In line with this, we can list various fields such as quantum computing and AI. However, before the presentation, I wondered if this would truly be effective. While the direction is correct, if the government were to focus on quantum computing, could we successfully develop quantum computing technology comparable to that of the US and China? Who can know that? Scientists? Bureaucrats? I believe it is very difficult. Ultimately, the market, in hindsight, is what knows, and the market makes the choice. When OpenAI first emerged, they aimed to build something based on GPT. At that time, GPT was based on Transformer technology, which was developed by Google. There were doubts about whether they could create GPT, surpassing a giant like Google, but $1 billion was invested.

including by Elon Musk. Eventually, under Sam Altman's leadership, GPT-3 was released and caught up with Google. Who knew that Sam Altman could surpass Google's Transformer technology with OpenAI? Nobody knew. However, the US made distributed investments in startups, and OpenAI succeeded among them. Other startups that made similar investments might have pursued different goals or given up. Only an ecosystem that allows for emergent challenges and failures in the market can create the core original technologies and advanced technologies that guarantee AI technology, as demonstrated by the US and China. No matter how hard we try with state-led initiatives, it will be difficult to enter the top 3 in AI technology. It is difficult to achieve results by following directions set by the state and universities. Moreover, the 100 trillion won we are considering is small compared to Amazon's AI budget of 145 trillion won.

Google or Apple would have even more. Therefore, in this money game as well, it is difficult for South Korea to compete with the US and China. So, what question should we ask now? When discussing AI development, AI-driven manufacturing innovation, and quantum computing, most people think of companies like Samsung, LG, SKT, and SK. Isn't it better for such companies to lead? However, in reality, the companies that have achieved major AI innovations in both the US and China are not global conglomerates. Intel is lagging behind along with Samsung, and Google and Microsoft are outsourcing their AI businesses. Why is this happening? These giant conglomerates have a size, organization, and inertia that are not suitable for pursuing disruptive innovations like AI.

Therefore, in an economy dominated solely by T-Rex-like large corporations, it is difficult to expect emergent innovation in AI. We need an ecosystem where smaller, more agile Velociraptor-like entities—in Jurassic Park, humans ultimately conquer not the T-Rex but the Velociraptors—can make emergent challenges, accept and collaborate with technologies needed by large corporations, and invest in promising companies early on to nurture them. Ultimately, disruptive innovation is more likely to come from outside than from within. While internal efforts are necessary, this is also confirmed by the failure cases of large companies in the US and China (e.g., Tsinghua Unigroup, Apple). Therefore, we must, in one way or another,

create an emergent ecosystem that suits us. Who should lead this emergent ecosystem? When Mark Zuckerberg founded Facebook, he was 19 years old. He is now in his mid-80s. In the movie 'The Social Network,' Mark Zuckerberg, facing difficulties due to lawsuits, runs out of funds and is cornered. At that time, John Parker, born in '79, who made a fortune in the P2P business with Napster during the first dot-com bubble, came to his rescue with his technological acumen and capital. Most importantly, he knew many venture capitalists who were willing to invest. John Parker saw Mark Zuckerberg's potential, led the investment, and eventually became Facebook's first president. He accumulated immense wealth. Most of the entrepreneurs leading the current revolution, including social media, are

born in the mid-80s. Yang Won-pyeong is also from '85. He was the developer of DeepC. The main players leading the current AI revolution are mostly from the late 90s, born in '97, which makes them 28-29 years old in Korean age. Meta is trying to recruit top-tier talent by offering them up to $100 million to $200 million, or 140 to 280 billion won. At Meta, Alex Wang, who leads the team, is also from '97 and signed a contract worth 20 trillion won. Lufri, one of the key engineers who developed DeepMind, is also from '95. This means that young talents in their late twenties are leading to innovative disruptive innovations in AI through relentless work, taking on immense risks, and expecting corresponding rewards.

We need to reflect on where our '97-born individuals are now, where the '85-born individuals who can guide them are, and where the '70s-born individuals who can bridge the gap for them are. Those with backgrounds in economics or political science have a solid theoretical foundation for understanding why innovation is difficult within large corporations. They know very well from historical examples that innovation comes from entrepreneurs who are small, agile, and risk everything in a situation where rewards are clearly promised. However, I believe it is a miscalculation that in Korea, people think state or large corporate leadership will suffice. According to the 2019 Nature paper 'Big team developed, small team disrupted,'

after analyzing 65 million patents and papers over 50 years, large teams primarily led incremental advancements, while small teams led radical innovations. This is academically validated and confirmed by cases in the US and Europe. Ultimately, if South Korea lacks an environment where vibrant individuals in their late twenties gather for all-night development, and if there are no venture capitalists who trust and invest in such young people for 5 to 10 years, is it truly possible for South Korea to achieve AI innovation and enter the top 3? Is it appropriate for the state to dictate the direction of quantum computing and AI? This is a difficult question.

Comparison of US and China AI Ecosystem Models

The US has a robust ecosystem centered around its vast capital market, with a strong foundation of private startups and well-established networks like Silicon Valley, allowing startups to spread effectively across key regions. China is leveraging its central-local government relationship to achieve quantitative growth through state-led initiatives and private collaboration. France and the European Union are adopting a strategy of securing technological sovereignty and stabilizing the European market through state-led initiatives. Looking at AI investment trends in the US, among the A, B, and C stages of venture investment in 2024, Stage A is the earliest, where investments are made based solely on an idea, with significant rewards for success. Stage B represents a certain level of achievement, and Stage C signifies a product that has received market response. Should the state guarantee all funding at Stage A across all fields, or at Stage C? This must be precisely designed to encourage adventurous investment while minimizing moral hazard, and it varies by field. In the AI sector, US venture investment is concentrated in AI. This is a strength of the US, but it can also be a weakness. When offering annual salaries of $100 million, who would consider applying AI to manufacturing? All AI developers will flock to Silicon Valley. This is both an advantage and a significant disadvantage of the US market. Political and economic experts point out that while talent poaching can create fair compensation, excessive poaching can undermine the industrial base. As you can see, venture investment in the US is very fragile. Therefore, at each stage of venture investment, the state must precisely design the level and type of compensation

to allow for bold investments while minimizing moral hazard. This is how the US ecosystem is structured. During China's deep learning revolution, innovations did not first emerge from companies like Baidu, Tencent, or Huawei, but from DeepMind. In 2015, they made money by establishing the High Flyer investment hedge fund. What were other companies doing at that time? It was the strength of small venture companies that were lightweight, miniaturized, and had enormous reward systems. Thus, DeepMind innovated rapidly and developed innovative products in 2023-2024. Behind this was strong support from the Chinese central and local governments, as well as Chinese conglomerates and investors.

to ensure that bold investments can be made while minimizing moral hazard. This is how the ecosystem is structured in the US. During China's deep learning revolution, innovations did not first emerge from companies like Baidu, Tencent, or Huawei, but from DeepMind. In 2015, they made money by establishing the High Flyer investment hedge fund. What were other companies doing at that time? It was the strength of small venture companies that were lightweight, miniaturized, and had enormous reward systems. Thus, DeepMind innovated rapidly and developed innovative products in 2023-2024. Behind this was strong support from the Chinese central and local governments, as well as Chinese conglomerates and investors.

The US is privately led and centered around its capital market, but having abundant capital and strengths does not solve everything. Excessive competition among companies leads to talent poaching, and incentives are concentrated upstream (in the most profitable and certain areas of value addition), resulting in a lack of incentives for midstream or downstream AI applications. I believe this is a crucial strategic target area for South Korea. Furthermore, industrial diffusion is insufficient, and although I hesitate to say this about the US, political instability can also be a risk factor. The financial market can easily overheat, leading to overpayment or excessive bonuses. In China's case, there is still a possibility of moral hazard due to the subsidy-centric structure.

Korea's AI Ecosystem Strengths and Strategic Target Areas

There is a risk of fragmentation from global standards, and Chinese technology may not become global technology. As US-China tensions escalate, this possibility increases. France also faces limitations with state-intervention-focused investment. South Korea not only has weaknesses but also relative strengths. It possesses a world-class semiconductor ecosystem, a strong manufacturing base with global manufacturing IT companies, and a high demand for digital transformation. It also has strategic planning capabilities within the government. It has experience in state-led growth, overcoming the foreign exchange crisis, and leading the restructuring of large corporations—in other words, 'muscle memory.' If the democratically elected government effectively utilizes this muscle memory, it can achieve success in AI transformation. Finally, it has an excellent talent pool and a Korean diaspora. Just as Chinese Americans have become key recruitment targets for Meta, there are many outstanding talents among Korean Americans and other foreigners, and this network can greatly contribute to AI transformation. Our analysis of publicly available data from OpenAlex, which assesses AI-related research capabilities, shows that South Korea's 2023 ranking is not bad. If you look at the country that has risen sharply there, it is Saudi Arabia, and due to the focus on the location of research institutions,

Just as Chinese Americans have become key recruitment targets for Meta, there are many outstanding talents among Korean Americans and other foreigners, and this network can greatly contribute to AI transformation. Our analysis of publicly available data from OpenAlex, which assesses AI-related research capabilities, shows that South Korea's 2023 ranking is not bad. If you look at the country that has risen sharply there, it is Saudi Arabia, and due to the focus on the location of research institutions,

Just as Chinese Americans have become key recruitment targets for Meta, there are many outstanding talents among Korean Americans and other foreigners, and this network can greatly contribute to AI transformation. Our analysis of publicly available data from OpenAlex, which assesses AI-related research capabilities, shows that South Korea's 2023 ranking is not bad. If you look at the country that has risen sharply there, it is Saudi Arabia, and due to the focus on the location of research institutions,

Saudi Arabia's ranking has sharply increased due to its intensive AI investments. This pertains to AI-related research papers. 2015 can be considered the starting point of deep learning; the company was founded in 2015, and deep learning began around 2016. At this time, we can see a tremendous quantitative increase in AI research from China (purple). An interesting question is whether this quantitative research has also led to qualitative research. The simplest way to check qualitative research is to look at citation indices. Looking at the top 1% citation index, although there may be a high possibility of co-citation within China, a significant change began to occur in China in 2015. The quantitative investments China had been making were now beginning to lead qualitative changes, and coincidentally, 2015 aligns closely with the development of initial models around 2015-2016, the starting point of deep learning. Therefore, the strategic starting point is precisely

South Korea needs to design a structure that allows for success within its constraints, rather than merely imitating the US and Chinese models. The state should act as an infrastructure provider and incentive designer. Instead of selective subsidies, universal subsidies should be provided, allowing even foreign companies investing in Korea to benefit from these subsidies. It is almost time for us to open our doors and attract foreign investment. Therefore, let us create a third way, a new model for fostering an AI ecosystem, neither Chinese nor American, where the government plays a nurturing role. And through the establishment of this ecosystem, K-pop is already expanding globally. After Seo Taiji's disruptive innovation, entertainment agencies initially emerged, but through controversies like slave contracts, fair distribution to artists, and even recently to choreographers, such reward systems have indeed brought about innovation in K-pop. In a similar vein, it is entirely possible for us to achieve large-scale investment, talent development, fair compensation, and foster autonomy and creativity. Therefore, the government should focus on infrastructure and

system design and promotion of the ecosystem. The five key players—large corporations, startups, VCs, university research institutions, and the government—each have their roles in creating the ecosystem. Coordinating these roles will likely fall to the government for the time being. Of course, as venture capital matures, the direction should shift towards VC leadership, with the government acting as an infrastructure provider, market cultivator, and system designer. My idea is that just as general trading companies played a role in aggregating issues for companies and providing financing and overseas market information when our companies were not yet developed, the state can now play such a role for startups that are just beginning. This includes providing tax benefits, corporate tax benefits like in the Guro Digital Complex, land compensation, or shared office spaces. And this is very important.

Finally, I would like to emphasize one last point. One of the most crucial aspects of specific policy proposals, and I heard that Professor Kwon Seok-jun is the only one from an engineering background here today, is the need to improve the treatment of engineers. If the university compensation system cannot be drastically changed, the government needs to develop a new reward system. This could involve indirect support through awards for outstanding engineers, with the government taking the lead in attracting world-renowned talent, and most importantly, the social reward system needs to change.

How many engineers are featured in children's books? There are celebrities and politicians, but very few engineers. Role models like these are crucial for elementary school students. For example, one person, Denison, inspired countless elementary school students to become robot scientists. In China today, deep learning developers are becoming social heroes, and while their reward system might seem excessive, I believe it is quite important. Therefore, I would like to suggest that the material and social reward systems for engineers need to shift significantly towards engineering, away from the current emphasis on Nobel Prizes or humanities.

Economic Security and Diplomatic Strategy in an Era of Complex Crises

Thank you. >> This is Professor Kim Hyun-chul from Seoul National University. We are about to begin the third session. As you can see from the overall title of today's event, 'Diplomatic and Security Strategy Directions for the Republic of Korea,' a traditional seminar would not require this third session. This is because the economy, industry, and technology discussed in this third session are separate topics with little relation to diplomatic and security strategy.

However, as you are all aware, we are currently in a situation where the economy and diplomatic security are inseparable and intertwined, and even technology is entangled with diplomatic security. This situation itself is a complex crisis. In the past, diplomatic security, economy, and technology would be discussed and addressed separately. But today, the economy, industry, and even technology must be considered together within the framework of diplomatic security, making this a crisis situation, hence the existence of this third session. And specifically, while our discussants have been assigned to other sessions, the economic domain is not just about economics but encompasses a wide range of aspects including industry and technology, which is why we have invited two discussants today.

First, the discussants, as you will soon realize from their discussions, will be providing commentary on the presentations. I will allocate 10 minutes to Professor Kim Yang-hee and 10 minutes to Professor Bae Hyung-ja, respectively, and I would like to ask them to provide their discussions or preliminary presentations from their unique perspectives.

Deepening Protectionism and US-China Hegemonic Competition

>> Yes. Hello. I am Professor Kim Yang-hee from Daegu University. I am honored to be a discussant for the presentations of two individuals from whom I have learned so much and whom I admire. I came here with the thought that perhaps I might be presenting again. Frankly, I am a bit sleepy. I had a radio interview at 7:30 this morning, so I've been quite preoccupied for the past few days. I cannot help but mention this in relation to the US tariff negotiations, as there were reports this morning that a resolution was expected, and I even went to the airport only to be told not to come, so I had to turn back. I've been receiving calls continuously. I've had a very hectic few days. I must address how this relates to the ongoing tariff war and how it directly connects to the presentations made by the two speakers earlier. I started my radio interview this morning by saying, 'The dam has broken.' Regarding the lack of agreement,

What I mean is, I myself was quite naive. Many of us, especially those involved in diplomacy, talk about solidarity among middle powers. And as mentioned in the previous session, there was talk of solidarity between Korea and Japan from an alliance perspective, but Japan betrayed us. Japan betrayed us. However, this is giving us a very significant turning point in the international order, and I feel that we were too hasty.

What this means is that Japan and the US won the battle. They have strangely created a win-win situation. I cannot go into detailed explanations, but that is my judgment. However, while both countries won the battle, I believe Japan has embarked on a path to lose the war. They may be pleased for now, having joined hands with the US and won, but if Japan, the EU, South Korea, Canada, and Mexico had held firm, it would have been a typical scenario. If they had held firm, the US could not have gone this far. However, what did the US consider? First, they needed to break the weakest link before these countries could join hands. Second, they needed to create a diversion domestically due to the 'time' issue. Frankly, I sensed that something would break a few days ago.

I personally believe, though it's my own opinion, that telling people not to come to Korea was due to the current severe 'time' issue. Why? Because it's not just anywhere else, but Maga is currently rising up. In that context, Korea alone is not enough. To focus on triggering something stronger in agreement with the EU, Korea is being pushed back for now. I sometimes speculate if this is the case. What I want to say is our perspective on the trade war. It's simply that what I find most difficult is this: when reporters call, they always ask, 'What should Korea do?' But every time, I say, 'Let's think about it one more time. What is it? What should Korea do?' Before that, let's calmly and accurately assess what the U.S. is thinking right now.

Let's start from there. When we consider this, what we need to think about more is that the U.S. is bringing about tremendous changes in the world order, and when we consider that a turning point in Korea-U.S. relations is now before us, we cannot find an answer if we don't view the trade war from the perspective of how the world order will unfold and how Korea-U.S. relations should be rebuilt. In that sense, more than anything else, what is Trump doing right now? Many people say that the U.S. has changed compared to the past.

But I want to point out one thing that is being overlooked: how does Biden's protectionism differ from Trump's protectionism? I have a concept I created called 'protectionist camp consolidation.' That is, Biden, wisely judging that he cannot defend China alone with his declining hegemony, has brought in allies and partners. I call this protectionist camp consolidation. Now, I can't see the diagram, but in reality, it's not just the U.S. that is consolidating its protectionist camp; rather, China is doing it more diligently now, with the expansion of BRICS, for example. But has this protectionist camp consolidation ended with the era of Trump? I don't think so. Protectionist camp consolidation is moving into version 2.0. What is it? Biden wisely used both the stick and the carrot. What is the stick?

Still high tariffs. However, with the carrot of the IRA and the CHIPS Act, he has created incentives for allies and partners to join. But in Trump's case, it's like, 'Why do we need that? Is this food?' Ah, yes, it is food. Other than that, the stick is enough. He believes he can lead the protectionist camp sufficiently by just wielding high tariffs. But, if I may. But the important thing here is, could you show the table behind me? Interestingly, Pence recently said something important, which is not well highlighted. What did Pence say? He brought up the concept of a 'North American Fortress.' At least, he wants to make Canada and Mexico his allies.

Then there is another important point that is often overlooked: Pence also mentioned the 'Great China Encirclement Ring.' This is where tariffs come in. Tariffs are not simply about opening markets and saying, 'I'll live well on my own.' Ultimately, Trump intends to solidify his allies in the U.S.-China hegemonic competition. However, unlike Biden, he believes that the stick is sufficient, without the need for a carrot. Ultimately, Korea, Japan, and typically the Philippines, as we saw, recently faced a choice: break up with China or not? If they break up, tariffs are reduced. If they don't break up? Tariffs are increased. This is how things are progressing. This is what I see as protectionism 2.0. So why am I talking about this? When I connect this to the previous two presentations, what should Korea do? I feel this is an extremely difficult homework assignment.

Securing Strategic Autonomy for Korea within the U.S.-led Bloc

In a multipolar era, if we are to continue to lead in semiconductors, which are crucial, and become a global top-three player in AI, we inevitably have to go along with the U.S., which still possesses strong competitiveness. When asked which protectionist bloc to join, we must align with the U.S.-led bloc out of sheer practical necessity, not based on values or ideology. However, the U.S.'s current approach offers no carrots, only sticks. When the U.S. wields only sticks, it appears to be alienating its allies rather than consolidating the protectionist bloc, leading to statements like, 'This is not the U.S. we knew.' This presents us with a truly difficult homework assignment, which has been my recent concern. Can we realistically achieve a certain degree of autonomy from the U.S. within the U.S.-led bloc, which we are inevitably and unavoidably part of?

Personally, I believed Korea should join NATO. This is because while NATO is clearly influenced by the U.S., it is not solely composed of the U.S. Within NATO, by befriending non-U.S. member states, we can, ironically, build security cooperation partners with other countries besides the U.S., and within that framework, we must create strategic autonomy. I wonder if this is possible. When I consider investing in AI or semiconductors again, I personally believe we have no choice but to go with semiconductors because we are so strong in that area. However, China is currently ahead of Korea in AI, despite our strength in semiconductors. While the previous two speakers, Professor Kwon and Professor Park, suggested it might be possible, I question how far China's AI ecosystem can advance, especially with the U.S. blocking it.

I want to pose the question: is it truly possible? Furthermore, from Korea's perspective, what I consider most important is self-reliance in security and manufacturing. If we proceed in that direction, to what extent can we achieve the self-reliance we desire while remaining within the U.S.-led bloc? At the same time, as was mentioned earlier, we face a reality where we must distance ourselves from China, but also from the U.S. To what extent can we go down our chosen path? Another rather painful question is that we certainly need AI in manufacturing, but China is already so far ahead in manufacturing, excluding semiconductors and shipbuilding, that can we truly maintain competitiveness in manufacturing by pursuing AI in the same way as China? What path should we take in that regard?

Securing Manufacturing AI Competitiveness and Data Sovereignty

So, should we focus solely on semiconductors and shipbuilding from the very beginning, choosing and concentrating on these areas? Should we abandon the rest? Or, conversely, is it precisely because of this that we must elevate the remaining sectors through AI to maintain our manufacturing competitiveness against China? I want to pose these serious questions and hear your respective answers. Since time is running short, I will briefly touch upon one point that I believe definitely needs to be addressed. While you both mentioned it in passing, it's impossible to discuss AI development without discussing data.

The Line incident is significant in discussions of South Korea's AI development, concerning how data issues will be handled, and in that Japan exerted economic coercion on an ally.

The previous administration dealt passively with the Line incident, deeming it a private company matter, and lacked a conceptual understanding of AI and data issues. Furthermore, although Line was one of the few leverage points South Korea could have against Japan in terms of economic security, Naver currently exercises little influence over Line. Therefore, if the new administration aims to become an AI powerhouse, a thorough re-examination of this issue is necessary.

South Korea's Diplomatic Choices Amidst US-China AI Competition

I have listened carefully to the presentations by Professor Seokjun Kwon and Professor Jonghee Park. I would like to offer a few questions and comments regarding the theme of US-China AI competition and the evolution of the AI industry. Professor Kwon mentioned that starting with China's challenge, the US is striving for AI leadership, and the AI industry structure is evolving uncertainly and dynamically. I am particularly focusing on China's challenge.

After deep learning, it was thought that the US was unilaterally ahead, but as the possibility of creating a China-led AI ecosystem emerged, attention began to focus on China. China is lagging behind the US in many aspects, including military power, the dollar, and soft power, and is trying to compensate for this through AI, investing heavily. As a result, achievements like deep learning have been made.

The US is checking China in three ways, including export controls on advanced process technologies. The first strategy, export controls, has succeeded in slowing down China's pace, but I believe it is a partial success and failure, given that China is catching up technologically through various means, as seen in deep learning.

China can build a China-led ecosystem by developing cost-effective models without massive capital investment. The second is manufacturing support, where the US aims to secure advanced process facilities domestically. I have doubts about whether the US can achieve 15% of the foundry market share by 2035, as projected in Professor Kwon's presentation. It is not easy to restore the US manufacturing ecosystem, which has been devastated over 30 years, through tariffs or subsidies in a rapidly changing environment, including the Trump administration's tariff policies.

The US is pursuing foundry business by attracting Taiwanese and South Korean companies, but mass production is expected around 2025, and there are issues such as cost increases and investment uncertainties. The Trump administration's subsidy policies can also be a variable. Therefore, it appears difficult for the US to achieve its target market share of 15%.

While the US's dominance will not be eroded, China's challenge will continue. In this complex situation, South Korea must make diplomatic choices such as self-reliance based on the ROK-US alliance, strategic autonomy, and balancing. In the technological field, we must question whether cooperating only with the US is the right choice for South Korea's future, given the uncertainty of cooperation with the US. I also want to emphasize the importance of technological diplomacy, along with the government's efforts to invest in AI and establish systems.

Regarding government investment in AI, looking at the past development of semiconductors and the establishment of internet infrastructure, large-scale investment is necessary in the current AI era. Semiconductor development in the 1980s was led by Samsung with government support, and internet infrastructure investment in the late 1990s was on a scale equivalent to 10% of the government budget at the time, enabling companies like Naver and Daum Kakao to grow.

Successful venture companies can emerge in the current AI era only with similar large-scale investments. However, the expansion and commercialization of AI infrastructure have not yet materialized, and the spread is slow due to the monetization of AI app usage. Therefore, scaling up AI infrastructure investment and the government's role are crucial.

The ecosystem creation mentioned by Professor Park Jong-hee is the role of companies. Similar to Samsung's semiconductor ecosystem development, the focus should be on infrastructure, human resources, systems, and legal aspects, rather than completing the supply chain. Finally, the technological diplomacy part is important. AI technology development cannot happen in isolation; the internalization of advanced technologies is essential.

Diplomacy is crucial for the advancement of AI technology. South Korea is in a situation where it has no choice but to form a technological community with the US, but this entails significant costs. The US does not easily transfer technology, and managing cooperation with China is also an important task. Responding to China's rapid technological catch-up while creating space for cooperation is a challenge.

Regarding AI Safeguards, recently mentioned by Naver's AI executive, the approaches at the corporate and national levels must differ. Like the 'Galapagos phenomenon' in the US IT industry in the 1990s, it is concerning that the discussion on AI safeguards is solely focused on defensive aspects. It must be balanced with public data and an open ecosystem.

In response to the US-led Big Tech AI ecosystem and China's authoritarian AI model, South Korea must develop a third AI model that embodies content, diversity, and democratic values. Expanding this in solidarity with middle powers and the Global South will be South Korea's overarching direction.

China's Technological Self-Sufficiency and the Limits of AI Development

I would like to ask both presenters to respond to the discussants' questions within approximately 4 to 5 minutes. Professor Seokjun Kwon will answer the questions first. The question regarding China's self-sufficiency in semiconductors and AI and its ability to overcome technological sanctions is very timely and important. China's innovation is driven by the industrial policies of the Chinese Communist Party government, but it would not have developed to this extent without US technological sanctions. This can be seen as 'innovation driven by scarcity.'

US sanctions strengthen internal opinion control in China and enable support for government policies. While top-down policies are effective in the early stages of an industry, bureaucracy and a lack of flexibility can arise in the mature stage. As Professor Park Jong-hee pointed out, if too much information is involved, it becomes difficult to keep up with changes.

China's policy is a mix of top-down, bottom-up, and mid-up approaches. In particular, state-owned enterprises under local governments create public funds and collaborate with private experts to create hybrid policies. This funding structure, combined with US sanctions, leads China to invest heavily in technological development for self-reliance. In addition to deep learning, attention should be paid to companies like Huawei.

Huawei is a company that spans almost all fields, starting from telecommunications equipment to semiconductor design and AI model development. With its self-designed chips and AI models, it is achieving performance comparable to Google's Gemini 2.5. This technological self-sufficiency can be seen as the US inadvertently giving wings to China.

However, China's technological self-sufficiency is a double-edged sword. While it may be efficient for the domestic market, it faces difficulties in expanding into overseas markets due to controls by the US and its allies. Its influence in the Belt and Road Initiative and among the Global South alone has limited ripple effects. Furthermore, China's policy-making process lacks democratic governance and is deficient in transparency, fairness, and rationality, which limits its growth.

China's AI development is innovation driven by scarcity, and US sanctions are actually promoting China's self-reliance efforts. While the technological advancements of companies like Huawei are noteworthy, China's technological self-sufficiency has limitations in the international market. Moreover, the lack of democratic governance will be a constraint on China's AI development.

US Foundry Market Outlook and Cost Issues

It was assessed that China is rapidly expanding its semiconductor manufacturing capabilities and is striving to overcome technological limitations by building a foundry ecosystem, fostering domestic equipment, and establishing an AI highway. Domestic production in the US is expected to reach over 15%, and possibly up to 20%. The problem is cost. The production cost of fabs operating in the US is inevitably at least 40% higher than that of local fabs in Taiwan.

The Importance of Data Sovereignty for Securing AI Sovereignty

With a 1.4 times higher cost, who would buy it? Therefore, the US government may likely guide policies to encourage major US foundry companies to purchase wafers produced in the US, similar to the Chinese government's approach. This is the answer to the second question. Let's talk about South Korea's AI ecosystem, particularly sovereignty. The current administration has put 'AI sovereignty' as a core state philosophy. Many people worry if this will lead to a Galapagos syndrome if it's only applicable in Korea. However, we can think of it this way. AI as we know it, especially AI based on large language models (LLMs), is not just simple software or chatbots.

It is like an operating system (OS) that can run a country. Through this OS, many things can be done. Narrowly speaking, optimization must be done using AI in the process of upgrading national institutions. However, we cannot entrust this entirely to the US or China. Especially, government systems cannot be entrusted to foreign OS. Earlier, you mentioned LINE, Mr. Kim. One of the reasons the Japanese government wanted to take over LINE so badly was not only because it contained personal information of nearly 100 million Japanese citizens, but more importantly, because they knew that LINE contained the largest amount of digital information in Japanese.

They wanted to create Japanese-language-based AI through this. However, if Naver controlled this, it would not proceed as the Japanese government intended, so they took it over semi-forcibly. From this, we can understand the importance of data sovereignty, which serves as the foundation for creating AI sovereignty. From this perspective, I believe that South Korea must have areas of security that are not part of the Galapagos syndrome, and we must sufficiently create an ecosystem that can guarantee sovereignty in those red lines.

China's Sovereign AI Ecosystem and South Korea's Opportunity

Earlier, I mentioned Huawei. Huawei could not have reached its current position solely with its own funds; it was due to the Chinese government's policy support. Huawei not only has its own Ascend chips but also distributes these chips to AI model development startups in China. The Chinese government guarantees these distributions, builds AI data servers across China, and mandates the use of Huawei chips instead of Nvidia chips in these servers.

This allows for testing and obtaining update data for the development of the next chip. This appears to be the formation of a kind of China-led sovereign AI ecosystem. South Korea can do better. While importing Nvidia chips is possible, there are various MP chips currently under development domestically. In the past, the focus was solely on developing large language models (LLMs), but now, small language models (SLMs) applicable to specific industrial domains are becoming more important. In this regard, an ecosystem is needed to create specialized chips developed in Korea. The companies that make these chips always say the same thing: they need to produce enough chips to move to the next stage, which requires investing hundreds of billions of won in fabs. However, most startups do not have that much capital and only have tens of billions of won. If one venture fails, most of them go bankrupt.

South Korea's Technological Diplomacy Strengths and Pilot Projects

Therefore, we are asking the government to provide guarantees, perhaps once or twice. And there is a need to establish a minimum infrastructure that can encompass these elements. The last question is about technological diplomacy. We have several strengths in technological diplomacy: democratic governance, global-standard commercial law, shareholder capitalism, and a diverse portfolio of technological capabilities and manufacturing. Not many countries possess these.

Germany, France, Japan, and South Korea are examples, but Japan is facing difficulties in the semiconductor industry, and Germany and France have industrial bases but are distant from advanced industries. It is no exaggeration to say that only South Korea possesses this level of industrial competitiveness. If democratic governance is guaranteed and we demonstrate transparent, reliable, and reproducible models within global frameworks, other countries can follow suit. Through this, we can export models, and when the US reshores manufacturing, it can use South Korea as a case study.

If many private companies find it difficult to undertake this, the government can implement pilot projects. The government can utilize institutions with the largest data holdings, such as the Health Insurance Review and Assessment Service. This institution has digitized decades of health data accumulated from 50 million people. By utilizing this data, various pilot projects can be carried out in areas such as AI in bio and AI in healthcare.

Furthermore, it is time for infrastructure updates. Following semiconductor and ADSL infrastructure, the next stage of infrastructure is becoming crucial. AI can be utilized in the process of optimizing energy, telecommunications, and more. This will serve as an excellent pilot case and can lead to self-sustaining AI ecosystems in each domain. Through this, we can create and export excellent models and lead global standards.

The Line Incident and Data Sovereignty Issues

I will conclude my remarks here. I often present after Professor Seokjun Kwon in discussions, and I am always so engrossed in his words that I forget I am supposed to be presenting. I just realized now that I am supposed to be presenting. The questions are so good that I don't have the answers, but I share the same concerns. What would have been the appropriate response from the South Korean government to the Line incident? What exactly were the Japanese government's response and the unfolding of the Line incident? Protecting our companies would have been the primary goal, so what was Naver's stance? Why is the discussion of AI sovereignty becoming a double-edged sword in the context of data sovereignty? We must be cautious in our wording and strategy because Japan might wield it against us.

How could Japan have been persuaded that the data built in Japan is held in South Korean data centers? What would have happened if there had been backlash against Naver in Japan if our government had taken a strong stance? These are important points to consider comprehensively. As more information emerges, it is increasingly likely to become a complex challenge for our companies and government. Furthermore, when exporting AI technology to regions like the Middle East, conflicts may arise regarding intellectual property rights, ownership, and processing issues.

The Role of the State and Survival Strategies in the Era of AI Transformation

That is a very important point. Your point, Ye Jae, about how we can create the changes that emerged during the growth from 4 million to 10 million users is also very difficult. What is the role of the state in creating an ecosystem, and how should it be done? I also want to know the answer to this frustrating question. You mentioned that there is no role like a solicitous but not overly strict parent who is liberal, open, and can step back at any time, unlike a general trading company that forms excessive partnerships. We need to find such roles. Ultimately, if we cannot find them, it is difficult to create something from nothing. If those born in the 60s created Microsoft and Apple, those in the 70s led the dot-com bubble, those in the 80s led the smartphone and social media revolution, and those in the 90s are now leading the AI revolution, then in the next 10 years, changes will occur as AI spreads to manufacturing, robotics, biotechnology, and all aspects of life.

I am confident that AI will become a general-purpose technology like electricity in almost all fields. For the state to plan and create an ecosystem for AI transformation is not a matter of choice, but rather a matter of not being swept away and finding survival strategies. Although the foundation is lacking and the role of the state is difficult, if we fall behind in the emergence of GPT, the most important development of the 21st century, it may be difficult to replicate the experience of emerging as the world's top two economic powers. Ultimately, it is not a matter of choice but a problem that we must all contemplate and find answers to. The fact that there is no given answer is never a reason to give up.

I would like to offer some general remarks. Although it is detailed on the slide, there are various roles the state can consider. In terms of infrastructure, there are issues such as GPU pooling, data centers, power, and water supply. In finance, we must consider support methods that can foster venture capital without causing moral hazard. For human resources, a plan is needed to effectively support the transition to AI and manufacturing by connecting elementary, middle, and high schools, universities, and vocational training. We need to brainstorm various ideas, such as tax systems and regional cluster strategies, and adopt a triage strategy that focuses on short-term, achievable actions that can yield results. In the absence of a blueprint, a piecemeal triage strategy must be implemented.

Establishing Relations with China and Technological Cooperation

I will only say that much. Although there are continuous questions and I feel like going for another round, Professor Seokjun Kwon wrote 'The Three Kingdoms of Semiconductors,' so I will ask only the final question. Please briefly explain how we should approach China.

First, I would like to state that I am not a China expert. In terms of security, we have no choice but to consider the US's perspective, and in terms of technology, we have no choice but to cooperate with the US. The US has demonstrated, since the late Biden administration, the possibility of imposing technological sanctions on various countries, including allies. In January 2025, the US Department of Commerce categorized exportable AI technologies into three tiers (Tier 1, 2, 3). Tier 1 included major allies such as South Korea and Japan, while Tier 3 included traditional adversaries of the US. Interestingly, Taiwan was included in Tier 2, indicating that the US can adjust its tiering at any time, even though Taiwan is not a formal ally of the US but has a very high degree of dependence.

Furthermore, there is the Foreign Direct Product Rule (FDP). Even if a product is manufactured abroad, if US technology is included even slightly in the production process, the US can be involved in the export of that product. This is close to an export licensing system or regulation. If South Korea were to export HB to China utilizing this, it would fall under the FDP regulations. Therefore, as long as the US holds the choke points for key technologies, we cannot escape this.

So, what kind of relationship should South Korea have with China? It would be ideal if we could maintain a balance as we do now, but the US government will increasingly and overtly force decoupling from China. In this case, most industries abandoned by China have already been surpassed or overtaken by China based on economies of scale, making direct competition difficult. Conversely, it is important to secure South Korea's position as the top option to replace China in industries that the US is sanctioning and restricting. I would like to make a bold proposal.

In China, fierce competition and high unemployment rates exist as shadows of industrial development. Even graduates from good universities find it difficult to get jobs. In the case of Huawei, they even lay off employees over 35 years old. Therefore, I believe it is important for South Korea to adopt a forward-looking attitude to allow bright Chinese youths to work in Korea. Of course, thorough security filtering will be necessary, but we should consider ways to utilize bright individuals who cannot go to the US or Japan by having them come to Korea and serve Korea's interests. Furthermore, it is important to continue cooperation with China within the scope of not crossing the US's security red lines.

There is a need. One point I particularly wanted to make, although it was not mentioned today, is that as we transition from 2023 to 2024, China, based on the Nature Index, has begun to surpass the United States in both the quantity and quality of basic science. This gap will widen in the future, as most of the US federal government's R&D programs have been significantly cut. Therefore, we need to create areas for cooperation in basic science, which China may lead, with the assurance that these will not yet be applied industrially in Korea. Furthermore, we need to proactively foster and attract talent by creating a larger pool, so that scientists leaving the US can come to Korea instead of China.

Economic Security and Korea's Response in an Era of Complex Crises

Due to time constraints, Professor Kwon Suk-joon spoke at a very rapid pace. As you may have gathered from the discussion, you are well aware of how important economics, technology, and industry are to foreign policy and security strategy. While today's discussion primarily focused on AI and semiconductors, we are not solely considering these two industrial strategies. Professor Kim Jong-hee emphasized the need to find our own path, while Professor Kwon Suk-joon pointed out the importance of observing trends in the US and China. Professor Kim Yang-hee highlighted the need to consider the shift from free trade to protectionism, and Professor Bae Jae stressed the necessity of technological diplomacy.

The discussants in Sessions 1 and 2 acknowledged the importance of issues such as the US military withdrawal, the Taiwan conflict, and China's containment. However, they also urged us to consider the impact on our economy, industry, and businesses amidst these circumstances. This is precisely what economic security entails, though it was omitted from today's discussion. It seems we are entering an era where economic and security issues are no longer viewed separately, but rather as interconnected complexities. Therefore, I am deeply grateful to the session chair for including economic security as the final session at the academy, and we will now conclude our session. Thank you.

Changes in the AI and Semiconductor Technology Security Landscape and Korea's Response Strategy

Changes in the Technology Security Landscape due to [AI + Semiconductors] and Korea's Response Strategy

US-China AI Hegemony Competition and Korea's Economic Security

Kwon Suk-joon: In the third session, I will discuss the competitive relationships between countries regarding advanced industries, particularly focusing on Korea, the United States, and China. As you know, traditional geopolitical logic is evolving, with the focus shifting towards semiconductors and, more recently, AI.

Notably, the current administration places significant emphasis on 'sovereign AI,' or AI sovereignty, as a core national agenda item. However, it is crucial to examine how this strategy can open new breakthroughs for industry and how the global competition for AI, driven by hegemonic ambitions in countries like the US and China, can be understood within the context of international politics, especially economic security. This is what I aim to discuss.

As you are aware, China poses the greatest threat to South Korea's semiconductor industry. Until a few years ago, China primarily focused on producing low-value-added semiconductor chips, but recently, a clear transition towards quality improvement is evident. Particularly in AI semiconductors, China is nearing self-sufficiency in its semiconductor supply chain. This not only has a profound impact on Korea's semiconductor industry but also signifies that many of the technological sanctions imposed by the US on China, especially those aimed at preventing the production of core semiconductor chips, are either ineffective or are being circumvented by China with developing strategies to counter them.

In China, the semiconductor industry's development appears to be synchronized with the five-year cycles of Xi Jinping's first, second, and third terms. From the first Semiconductor Big Fund in 2014 to the second in 2019 and the recent third fund in 2024, not only has the scale of these funds increased, but the third phase is characterized by the comprehensive application of semiconductor technology not just in large-scale industries but also in AI semiconductors, mobility, power, telecommunications, and even energy and advanced biotechnology. One of the key emphases during this year's Two Sessions was 'new quality productive forces,' which reflects China's strong commitment to producing high-quality, high-value-added semiconductors.

Within this landscape, several key companies form China's AI and semiconductor ecosystem. These include not only the well-known Huawei but also companies like DeepSeek, which recently shocked many countries including Korea, and Alibaba, which is developing powerful Large Multimodal Models (LMMs) comparable to OpenAI and Claude. From Korea's perspective, it is crucial to note China's intensive investment in the semiconductor industry, particularly in semiconductor manufacturing. Recent reports indicate that China has secured at least three foundry companies that could rank among the top ten globally in the production of system semiconductors, which are critical for foundry operations.

More interestingly, Taiwan's influence in the foundry sector has been gradually diminishing. While the dominance of key players like TSMC remains largely unshaken, the influence of supporting companies such as UMC, PSMC, and Vanguard is declining, with Chinese companies increasingly filling the void. I highlight this because, although Chinese foundries have not yet reached the technological level of TSMC or Samsung Foundry, they are significantly expanding the number of fabrication plants. Furthermore, the technological capabilities of these plants are also advancing.

Consequently, Chinese foundries are expected to gain increasing influence in various sectors beyond specific high-performance semiconductors like AI chips or Apple Silicon. This includes areas such as industrial semiconductors, biotechnology, and general power and communication semiconductors, which utilize relatively mid-tech or legacy processes. It is highly probable that within the next decade, they will account for over one-third of the global legacy foundry market. Moreover, it is important to note that these foundry companies are not solely focused on manufacturing. Foundries naturally require corresponding process equipment. Historically, China's weakness in process equipment was its high dependence on US, Japanese, and Dutch equipment for advanced processes.

However, this dependence is also being mitigated by triple subsidies for domestic equipment as foundries expand quantitatively. These triple subsidies function as follows: for instance, when a foundry is established, the Chinese government provides subsidies; when the foundry purchases Chinese-made equipment, subsidies are provided to the equipment manufacturer; and incentives are offered to companies purchasing the chips produced, creating a triple subsidy structure. While a significant technological gap with the US persists, we are observing a considerable reduction in these gaps. Furthermore, as the semiconductor foundation is laid, a natural linkage to the next stage, AI, is being observed.

When we discuss semiconductors and AI, we rarely discuss them in isolation; they are discussed together. This is because, no matter how well companies like NVIDIA design GPUs, a robust manufacturing ecosystem is essential to physically produce these designed chips. Without specialized foundry fabs like TSMC, even a company as prominent as NVIDIA can only design GPUs, not manufacture them. China is currently laying this groundwork and building a highway for greater AI internalization.

Taiwan plays a pivotal role in this process. Beyond its foundry capabilities, Taiwan holds significant dominance in the cutting-edge process nodes below 10nm. Consequently, the concept of a 'Silicon Triangle' involving Taiwan, the US, and China has been prominent. The US is actively pursuing strategies to mitigate its high dependence on Taiwan for advanced semiconductor production.

The first strategy involves reshoring key advanced process fabs from Taiwan to the US. For example, TSMC is investing $165 billion to build 3nm process fabs near Phoenix, Arizona. Samsung Electronics is also constructing over nine fabs in Taylor, Texas, with a high probability that many of these will employ advanced processes below 5nm. Nevertheless, the extent to which China can catch up in advanced processes within the Silicon Triangle of Taiwan, the US, and China will be a crucial factor determining economic security in the region and, by extension, security in the Indo-Pacific.

If current trends continue, the US will secure at least 20-30% of its supply chain by the mid-2030s, a significant increase from the current less than 5%. This will diminish Taiwan's dominance. Another factor to consider is the potential increase in China's sub-10nm fabs. Currently, objective figures show that China's sub-10nm foundry fabs account for only about 2-3%. However, if current investments proceed as planned, this proportion could reach at least 5% by the mid-2030s. Furthermore, Chinese foundries, particularly Huawei's shadow fabs, may strengthen their control over Taiwanese foundries ranked 2nd, 3rd, and 4th, which are experiencing weakened capital and competitiveness.

US AI Hegemony Strategy and Korea's Role

These developments have the potential to significantly alter the geopolitical landscape of the Silicon Triangle. The US is actively formulating numerous hegemonic strategies for AI from a national policy perspective. The Stargate Project, announced immediately after the commencement of the second Trump administration earlier this year, is a prime example of such US strategy. The Stargate Project is massive in scale.

It amounts to $500 billion. While private companies participate in a significant portion of this fund, the US government also makes direct and indirect investments. Furthermore, the White House Office of Science and Technology Policy's recent white paper, 'Winning the Race: America's AI Action Plan,' reveals more explicit US AI hegemony strategies.

I describe them as explicit because the US has already been implementing export controls on critical AI assets like GPUs to its allies and core stakeholders. The recently updated strategy suggests that the US may effectively compel key allies to participate in the US-led AI technology ecosystem. Simultaneously, it indicates a strengthening of control policies towards key competitors who will be denied access to this ecosystem, and as everyone knows, these key competitors refer to China. Therefore, access to the AI ecosystem, as defined by the US, ultimately becomes a battle over standards.

Naturally, when discussing these standards, we must consider not only AI models and ecosystems but also the subsequent stages. The 'subsequent stages' refer to all spillover effects as AI is applied to various industries, including manufacturing and defense, beyond its standalone applications. The US is currently discussing how 'manufacturing is weak' and 'reshoring is difficult,' but it appears they view AI as a game-changer. Notably, the US white paper suggests that regulatory and safety concerns regarding AI are being temporarily set aside. It emphasizes the importance of more active AI utilization, indicating a willingness to dismantle significant private sector regulations and maximize performance to pave the way for AI advancement. The semiconductor foundation is crucial for the US to maintain these AI hegemony strategies. Therefore, if Korea is to concretize its cooperation with the US in terms of security, particularly economic security, it should propose the following to the US: The US currently lacks the capacity for complete domestic production of advanced semiconductors and, consequently, must rely on East Asia, particularly Taiwan, on which it is most dependent. However, Taiwan is not a US ally and has no diplomatic relations.

It is important to utilize AI more actively. To this end, I believe they are demonstrating a willingness to remove a significant portion of regulations on the private sector and maximize performance to open a superhighway towards artificial intelligence. For the US to continue this AI hegemony strategy, a semiconductor foundation is crucial. Therefore, if Korea has a strategy to solidify cooperation with the US from the perspective of security, particularly economic security, Korea should propose the following to the US: The US currently lacks the capacity for full domestic production of advanced semiconductors, and consequently, must rely on East Asia, especially Taiwan, on which it currently depends the most. However, Taiwan is not an ally of the US and does not have diplomatic relations.

China's AI Technology Internalization and Challenges

Consequently, countries that can form alliances with the US and play a key role in the supply chain are Korea and Japan. However, Japan's semiconductor manufacturing base has significantly weakened. Therefore, from a technological standpoint, Korea can become the most crucial technological partner in the US's AI hegemony strategy, as the US cannot partner with China. This is something we can consider. So, what is China thinking? China's strategy is singular: to endure. It is gradually internalizing technologies in the semiconductor and AI fields where the US is imposing sanctions. Even with internalization efforts, a technological gap remains. While overall progress has reached about 70%, as seen in the DeepSeek shock in January, China seeks to circumvent US sanctions if overcoming them proves difficult. If circumvention is also challenging, they experiment with various destructive methods. They possess ample capital and an overwhelmingly large pool of specialized personnel to conduct these experiments. Approximately 70% of the top 100 AI research institutions globally are Chinese or of Chinese descent. This means that global AI technology has reached a level where it cannot function without Chinese talent. China is gaining confidence in this area.

It is important to note that key private companies in China's semiconductor and AI ecosystem, such as Huawei, Alibaba, ByteDance, Baidu, Longsoon, and SMIC, are rapidly internalizing their full-stack semiconductor and AI capabilities. These internalized technologies are transitioning from quantitative growth to qualitative growth. From Korea's perspective, it is necessary to consider how to utilize AI. While the conventional wisdom has been that scaling up models is crucial, the future focus will be on 'what inferences are possible' and 'what applications are feasible' within specific industries.

Building Korea's AI Ecosystem and Sovereign AI

Therefore, the development of purpose-oriented AI semiconductors suitable for specific tasks is becoming increasingly important, beyond expensive GPUs for building large-scale AI. The current monopolistic structure of the global AI and semiconductor supply chain, dominated by a few companies like NVIDIA, TSMC, SK Hynix, and Foxconn, is bound to diversify. This diversification may present a crisis, but it can also be a significant opportunity for Korean companies. Specifically, the next frontier for AI semiconductors lies in other manufacturing sectors.

This expansion can extend to other industries beyond semiconductors, including energy, biotechnology, shipbuilding, aerospace, and steel manufacturing. Korea, as one of the few developed nations capable of expanding its industrial base while maintaining democratic governance, can play a crucial role. To this end, Korea is pursuing sovereign AI and plans to establish AI data centers in various regions. The key challenge here is whether Korea can lead the infrastructure development necessary for such a level of AI investment.

While power grids, industrial water supply, and communication networks are vital, these are industrial policies that require a long-term perspective, not just a five- or ten-year government term. I always emphasize that these initiatives will require significantly more cost and time than projects like the Incheon Airport or KTX construction. Therefore, a long-term vision is essential for the implementation of such policies.

Finally, Korea needs to present case studies to the world demonstrating how AI can generate new revenue streams through its role in manufacturing, an area where Korea uniquely excels. I believe this can serve as a significant bargaining chip in negotiations with the United States.

Leap Strategy Through an AI Ecosystem-Centric Approach

Leap Strategy Through an AI Ecosystem-Centric Approach: Multiple Paths for State-Led AI Industrial Development

Leapfrog Strategy Through a Centralized AI Ecosystem Approach: Multiple Paths for State-Led AI Industry Development

Park Jong-hee: My presentation was intended to cover the broad theme of Korea's economic security, particularly the economic transformation centered on AI and how to create a vision for Korea amidst the US-China conflict. I will focus on delivering a single message through selective emphasis. For more detailed information, please refer to the economic security strategy report for the new government, which our Economic Security Cluster has submitted for printing. We would appreciate your reference upon its publication.

AI Technology Innovation Ecosystem: State-Led vs. Emergent Challenges

The current administration's AI pledges can be summarized as aiming for Korea to become one of the top three AI powers globally. Given that the US and China are likely to be the top two, Korea aims to follow them. The plan involves private investment of 100 trillion won, increased government budgets, and the establishment of a public fund if necessary. In terms of infrastructure, as Professor Kwon Suk-joon mentioned, 50,000 GPUs will be purchased for sharing, AI data clusters will be established in various regions to secure power supply, and comprehensive AI social infrastructure will be built. To foster talent, the establishment of AI-focused colleges and improved compensation are also being actively pursued. Additionally, various fields such as quantum computing and AI x can be considered. However, I have pondered this question.

Before this presentation, I questioned whether this direction would lead to success. If the government decides to focus on quantum computing, can Korea successfully develop quantum computing technology comparable to or on par with the US and China? Neither scientists nor bureaucrats can be certain. Ultimately, the market determines this. The market makes the choice. When OpenAI decided to build something based on GPT, GPT was a transformer technology developed by Google. People questioned whether it was possible to create GPT and surpass Google, a giant corporation. However, investors, including Elon Musk, invested $1 billion, and ultimately, under Sam Altman's leadership, GPT 1, 2, and 3 were released, catching up to Google.

Who knew that Sam Altman and OpenAI could surpass Google's transformer technology? No one knew. However, the US made diversified investments in startups, and OpenAI succeeded. Other startups that pursued similar development and investment may have gone in different directions or given up. Only emergent challenges and failures in the market, and the resulting ecosystem, can create the core foundational and advanced technologies that guarantee AI technology, as demonstrated by the US and China. Regardless of state-led initiatives, it is difficult to enter the top three AI powers without alternative approaches. Effective technological advancement cannot be expected through directions dictated by the state or universities.

Furthermore, while we consider a budget of 100 trillion won, Amazon's AI budget is 145 trillion won, and Google and Apple have even larger budgets. In this capital competition, it is difficult for Korea to compete with the US and China. So, how should we frame the question? When discussing AI development, manufacturing innovation, and quantum computing, we naturally think of large corporations like Samsung, LG, and SK. We consider whether it would be beneficial for them to lead. However, in both the US and China, the companies that have led major AI innovations are not global conglomerates. Intel is falling behind alongside Samsung, and Google and Microsoft are outsourcing their AI businesses. Why is this happening? Large corporations possess a scale, organizational structure, and inertia that are not conducive to disruptive innovations like AI.

Therefore, an economy dominated solely by large corporations, like Tyrannosaurus Rex, is unlikely to foster emergent AI innovation. We need entities like Velociraptors, smaller and more agile, alongside these giants. Just as in Jurassic Park, it is the Velociraptors, not the Tyrannosaurus Rex, that open doors locked by humans. We need a relationship where smaller dinosaurs like Velociraptors take emergent challenges, and their technologies are adopted and collaborated upon by large corporations. Furthermore, an ecosystem must be created where large corporations invest in promising companies from the early stages through venture capital. Ultimately, disruptive innovation is more likely to come from external sources than internal ones. While internal efforts are necessary, the failures of large corporations in the US and China (e.g., Tsinghua Unigroup, Apple) confirm this.

Ultimately, we must create an emergent ecosystem that suits us. Who should lead this emergent ecosystem? Mark Zuckerberg, the founder of Facebook, was 19 years old when he founded the company. In his 80s, he faced legal troubles, as depicted in 'The Social Network,' ran out of funds, and was cornered. At that time, Sean Parker came to his rescue. Born in '79, Sean Parker made a fortune in P2P business with Napster during the first dot-com bubble, possessing both technological acumen and capital. Most importantly, he knew many venture capitalists willing to invest. Sean Parker recognized Mark Zuckerberg's potential, facilitated investment, and eventually became Facebook's first president, accumulating immense wealth.

Most entrepreneurs leading the current SNS revolution are in their mid-80s. Yang Won-pyeong is also from '85. The main drivers of the AI revolution are primarily from the late 90s, specifically those born around '97 (28-29 years old in Korean age). You have seen in the news the salaries offered to these individuals when they receive job offers from Meta. Meta offers $100-200 million (140-280 billion won) to recruit top-tier talent. Alex Wang, the leader of this team, is also from '97 and has already secured a contract worth 20 trillion won. Ruffree, a key female engineer who participated in the development of DeepMind's model at the company founded by Yang Won-pyeong, was born in '95.

In other words, young talents in their late twenties, taking immense risks but expecting commensurate rewards, are driving innovative and disruptive AI innovations through their relentless work. It is time to reflect on where our '97-born generation is, where the '85-born generation who can lead them is, and where the '70s generation who can bridge the gap for the '85-born generation is. Those with backgrounds in economics or political science already possess the theoretical framework for why innovation is difficult within large corporations. History clearly shows that innovation originates from entrepreneurs who, in a small, agile, and secure environment with guaranteed rewards, commit everything. However, I believe it is a miscalculation to assume that this can be achieved through state or large corporate leadership in the AI field.

According to the 2019 Nature paper 'Big team develops, small team disrupts,' which analyzed 65 million patents and papers over 50 years, large teams drive incremental progress, while small teams lead radical innovation. This paper has been academically validated and confirmed by cases in the US and Europe. If Korea lacks an environment where vibrant young individuals in their late twenties gather and dedicate themselves to all-night development, and if there are no venture capitalists who encourage them to focus on development for 5-10 years with unwavering trust and investment, is it truly possible for Korea to achieve AI innovation and enter the top three AI powers? Is it right for the state to dictate the direction of quantum computing and AI x? This is a difficult question.

Comparison of AI Ecosystem Models in the US and China

Comparing the US and China, the US has a robust ecosystem centered on private startups within a vast capital market, supported by strong networks like Silicon Valley, with startups spreading across key regions. China is leveraging central-local government relationships, combining local government leadership with private sector cooperation to transition from quantitative to qualitative growth. France and the EU are pursuing strategies to protect technological sovereignty and stabilize the European market through state-led initiatives. Examining US AI investment trends in 2024, venture capital is divided into stages A, B, and C. Stage A is the earliest phase, where investment is made based on ideas alone with no certainty, but significant rewards are given for success. Stage B involves some level of achievement, and Stage C signifies a product that has achieved market traction.

Should the state fully fund and guarantee all areas, or provide support at specific stages? This requires meticulous consideration. The design must encourage risk-taking investment while minimizing moral hazard, and this varies by AI field. The US's strength lies in venture capital's concentration on AI, but as Professor Kwon Suk-joon pointed out, it can also be a weakness. When investing $100 million, who would think about AI applications in manufacturing? If you could develop AI, you would flock to Silicon Valley for a $100 million annual salary. This is the advantage of the US market, but also a significant drawback.

Poaching of talent in politics and economics can lead to fair compensation, but excessive poaching can undermine the industrial base. Venture capital in the US is quite fragile. Therefore, the state must precisely design the level and type of compensation at each stage of venture investment. The design must allow for bold investments while minimizing moral hazard. In China's case, when the DeepSeek revolution occurred, why did it emerge from DeepSeek rather than from Baidu, Tencent, or Huawei? In 2015, the High Flyer investment hedge fund was established and generated profits. What were other companies doing during that time? It was the strength of small venture companies that were lightweight, miniaturized, and offered substantial rewards. This allowed DeepSeek to innovate rapidly and develop groundbreaking products in 2023-2024. Behind this were not only the Chinese central and local governments but also strong support from Chinese large corporations and investors.

The US operates on a capital market-centric, private-led model. However, neither the US nor China possesses only strengths, nor does having ample funds solve all problems. Excessive competition among companies leads to talent poaching, and the incentive structure, which concentrates on the upstream with the most certain value and rewards, results in insufficient incentives for AI applications in the midstream or downstream. I believe this area represents a crucial strategic foothold for Korea. Furthermore, industrial diffusion is insufficient, and political instability, though I hesitate to say this about the US, can also be a risk factor.

Korea's AI Ecosystem Strengths and Strategic Focus Areas

Financial markets can easily become overheated, leading to overpayment or excessive bonuses. In China, moral hazard due to a subsidy-centric structure remains a possibility. If isolated from global standards and becoming a 'Galapagos,' Chinese technology may not become global technology, and this possibility increases as US-China tensions escalate. France also faces limitations with its state-intervention-focused investment approach. While Korea's weaknesses have been discussed, it also possesses numerous relative strengths. It has a world-class semiconductor ecosystem, a base of global manufacturing and IT companies, and a strong demand for digital transformation within its manufacturing sector. Additionally, it has strategic planning capabilities, evidenced by its experience in leading state-led growth, overcoming the foreign exchange crisis, and driving corporate restructuring (muscle memory).

If the democratically elected government effectively utilizes this 'muscle memory,' it can achieve success in AI transformation. Finally, Korea has an excellent talent pool and a Korean diaspora. While Chinese Americans have become key targets for Meta's recruitment, attracting significant attention, there are also a considerable number of Korean Americans and other talented foreign individuals, and their networks can greatly aid AI transformation. Our analysis of Open Alex indicates that Korea's ranking in AI-related research capabilities is not poor as of 2023. Observing countries with rapid ascents, Saudi Arabia's ranking has sharply increased due to its intensive AI investments.

This graph shows the number of AI-related research papers. 2015 was the year the High Flyer company was established, and DeepSeek's inception was around 2016. At this point, we can see a significant quantitative increase in China's (purple) AI research. An interesting question arises: did this quantitative research translate into qualitative research? A simple way to assess qualitative research is by examining citation indices. Looking at the top 1% citation index, although there is a possibility of increased co-citation within China, a significant change began to occur in China around 2015.

Recommendations for Building Korea's AI Ecosystem

China's quantitative investments have begun to drive qualitative change, coincidentally aligning with the period of 2015-2016 when DeepSeek's early models were being developed. Therefore, the strategic starting point for Korea should be to design a structure that enables success within its constraints, rather than merely imitating the US and Chinese models. The state should act as an infrastructure provider and incentive designer. Instead of selective subsidies, universal subsidies should be provided, and the doors should be opened to foreign investment by allowing even foreign companies investing in Korea to benefit, thereby attracting foreign investment. A new model for nurturing the AI ecosystem, distinct from Chinese or American approaches, should be created by the government.

Through the development of this indigenous ecosystem, similar to how K-pop has expanded globally, the emergence of entertainment agencies after Seo Taiji's disruptive innovation, and the subsequent controversies over exploitative contracts, have led to fair distribution for artists and even choreographers, ultimately driving K-pop innovation through its reward system. Similarly, large-scale investment, talent development, fair compensation, and the fostering of autonomy and creativity are entirely possible. Therefore, the government should focus on designing and encouraging the infrastructure, institutions, and ecosystem. The five key players—large corporations, startups, VCs, university research institutions, and the government—each have a role in creating the ecosystem. Coordinating these roles should be the government's responsibility for the time being. Of course, as venture capital matures, the approach should shift to VC leadership, with the government serving as an infrastructure provider, nurturing talent, and designing institutions.

Improving Compensation for Engineers and Social Reward Systems

I believe that just as general trading companies played a role in solving problems for companies and providing trade finance and overseas market information when our businesses were not yet developed, the state can play such a role for startups that are just beginning. The government can provide solutions such as tax benefits, corporate tax incentives like those in Guro Digital Complex, land compensation, and the creation of shared offices. This is very important. Finally, I will emphasize just one more thing. One of the most crucial aspects of concrete policy proposals is the improvement of treatment for engineers. If the university compensation system cannot be drastically changed, the government needs to develop a new reward system.

The government should provide indirect support through a system that rewards outstanding engineers and take the lead in attracting world-class S-class talent. Above all, the social reward system must change. How often do engineers appear in children's books like 'Why?' Celebrities and politicians are featured, but engineers are rarely mentioned. Role models are very important for elementary school students. For example, Dennis Hong alone has inspired countless elementary school students to aspire to become roboticists. Currently, in China, a deepfake developer is becoming a social hero, and while the reward system might seem somewhat excessive, I believe it is quite important. Therefore, I would like to state that a significant shift in focus is needed in both material and social reward systems towards engineers, moving away from the existing emphasis on Nobel Prizes or humanities towards an engineering-centric approach.

Speaker Introductions

Seokjun Kwon, Professor at Sungkyunkwan University.

Jonghee Park, Professor at Seoul National University.

Managed and Edited by: Inhwan Oh, Senior Researcher at EAI
Inquiries: 02 2277 1683 (ext. 202) | ihoh@eai.or.kr