[AI and New Civilization Standards Special Report] Economic Challenge ①: Decoupling of US-China Artificial Intelligence Ecosystems

I. US-China AI Competition and Differences in Paradigms

As Artificial Intelligence (AI) is expected to bring about revolutionary changes across science and technology, industry, and the economy, major countries worldwide are engaging in AI competition. AI competition is not a one-dimensional contest requiring superiority in various areas such as research and development (R&D), talent, infrastructure, and commercialization, but rather encompasses a multi-faceted competitive nature. This is why analyses and evaluations of the current state and prospects of AI competition include not only rankings of research institutions and companies, the entities driving AI technological innovation, but also national rankings that incorporate government strategies as a key evaluation factor.

The AI competition between the US and China is not limited to technological prowess. Implementation, innovation, and investment are known as the three major competitive domains of AI. Implementation encompasses talent, infrastructure, and operating environments; innovation covers research and development; and investment includes government strategies and private sector startups and investments. Following the US and China, Singapore, the UK (investment), and Germany form a third group, significantly trailing the US and China (<Table 1>).

<Table 1> AI Capabilities of Major Global Powers

Source: Tortoise Media. “The Global AI Index.”

Narrowing the competitive landscape to the US and China, the US shows consistent superiority in areas such as talent, infrastructure, and R&D, as well as commercial investment. The US advantage is particularly prominent in private investment. From 2013 to 2023, US private investment reached $335.2 billion, significantly outstripping China's $103.7 billion (Bomey 2024). The US also maintains a superior position in securing top AI talent. The proportion of top 20% AI researchers working in US institutions increased from 27% (2019) to 38% (2022), supporting this claim.

<Figure 1> US and China AI Investment Patterns

Source: World Economic Forum. 2020. “World order is going to be rocked by AI – this is how.” February 13.

Nevertheless, the prevailing view that US-China AI competition is intensifying is closely related to the formation of very different AI ecosystems in terms of AI commercialization and regulation between the two countries (Lazard 2023). While the US sees substantial capital investment centered on private funding, a cohesive national AI strategy is relatively weak. In contrast, China pursues government-led industrial policies, with government investment playing a overwhelmingly larger role compared to the US. China's government investment is reported to be 1.5 times greater than the combined public AI investments of all other countries.

The operating environment appears to be at a relatively lower level in the US. This is related to the US government's policy on AI regulation fluidly shifting between self-regulation and government oversight, and growing public concern about the expansion of AI utilization. In contrast, regarding regulation, the Chinese government is moving swiftly at the national strategy level, announcing new rules for AI developers and deployers in April 2023. It is clear that these rules fundamentally conflict with the US AI paradigm, as they mandate the implementation of core socialist values such as safeguarding national sovereignty and preventing regime subversion.

The gradual decrease in personnel movement between the US and China, and the increasing proportion of individuals staying within their own countries, also foreshadows reduced cooperation and intensified competition between the two nations. Specifically, 82.6% of AI talent in the US graduated from US graduate schools. Among AI talent who completed undergraduate studies in China and then pursued graduate studies in the US, the figure is 31.2%. Conversely, only 1.93% of US AI talent graduated from Chinese graduate schools. Notably, the proportion of US AI talent who completed both undergraduate and graduate studies in China is even lower at 1.54% (<Figure 2>).

In China's case, 75.5% of AI talent graduated from graduate schools in China, and 74.6% completed both undergraduate and graduate studies in China. While these figures are slightly lower than in the US, they indicate that China is nurturing its own AI talent. However, 8.8% of Chinese AI talent who graduated from undergraduate programs in China went on to US graduate schools, and 10.65% of those who graduated from US graduate schools were employed within China, suggesting a significant flow of AI talent to the US (<Figure 2>).

<Figure 2> AI Talent Mobility (2023)

Source: MacroPolo. 2023. “The Global AI Talent Tracker 2.0.”

II. US Strategy

AI's profound impact and its critical role in securing long-term future competitiveness make the competition between the US and China inevitable and fierce. The US is accumulating domestic capabilities to maintain and expand its technological lead over China, while simultaneously strengthening containment measures to slow down China's AI progress. The US is dedicating all its efforts to maintaining "as large of a lead as possible" over China (The White House 2022). Specifically, the US pursues a strategy to widen the technological gap with China through export controls on AI semiconductors, Electronic Design Automation (EDA) software, and military-use semiconductors. The general outlook is that if US export controls on China are strictly enforced, China's pursuit can be effectively contained. The US containment strategy is expanding to block the transfer of intangible technologies. The US preemptively prevents technological leakage through education and R&D, and is also adapting its talent recruitment from the perspective of technological leakage.

The US, in particular, views AI as a catalyst for China's military-civil fusion and is continuously increasing the intensity of export controls to fundamentally block access to core AI technologies. As it has become evident that "the Chinese military is the end-user of US semiconductors" (Weinstein 2021), concerns are growing within the US about the possibility of the Chinese military employing supercomputer simulations in its military operations. This was the background for the Biden administration's strengthening of export controls on semiconductors from Nvidia and AMD in August 2022.

The US strategy toward China can be described as a dual strategy: fundamentally blocking the conversion of US AI technology for military purposes while simultaneously enhancing the competitiveness of the US AI industry. This is because US export controls have a dual effect, impacting not only China's military-civil fusion but also the cloud services of major Chinese AI companies such as Baidu, Tencent, and Alibaba. The US government expects that if export controls are effectively maintained, it can delay China's AI progress, thereby maintaining or widening the current gap.

III. China's Strategy

From its early stages of development, China has pursued an AI development paradigm that is distinct from, and even conflicting with, that of the US. Since the announcement of the "Next Generation Artificial Intelligence Development Plan (一代人工智能发展规划)" by the Chinese State Council in 2017, AI technological innovation has been driven at the national strategy level. As evidenced by the Chinese government's advocacy of a "new national system for tackling key core technologies" (关键核心技术攻关新型举国体制), it has formalized its plan to overcome US containment through self-reliance and self-strengthening in AI technology. In this regard, China has identified 35 "choke point" technologies that could be used by the US for containment, and it assesses that it possesses the capability to counter US containment in 21 of these technologies ("EET China 2023).

Efforts to strengthen independent capabilities to counter US containment, such as export controls, are also evident in China's AI initiatives. China anticipates that the US AI strategy will increasingly intensify and expand its containment measures to delay or block China's progress, and therefore prioritizes building its independent capabilities. The Chinese government has concluded that since the US will continue its strategy of using key points in the AI value chain as choke points, the only way to overcome this is to strengthen domestic innovation capabilities. The Chinese government is focusing on financial support through special funds, various financial incentives, domestic semiconductor production policies utilizing the domestic market, and nurturing AI talent by linking industry and education systems. In particular, with weakened research exchanges with the US, a shortage of advanced talent is expected to worsen. The Chinese government plans to cultivate approximately 250,000 AI talents by 2024 and actively attract talent from countries other than the US.

A characteristic of China's AI strategy is the organic combination of government coordination and market-based innovation to form an independent AI ecosystem. The promotion of cooperation between national research institutes and leading companies is a prime example. Furthermore, the Chinese government has decided to proceed with mega-projects to enhance computing power as infrastructure, and as of March 2023, data centers are under construction in 30 cities. To support this institutionally, the Chinese government has established the National Data Bureau (国家数据局) and is pursuing a strategy that combines data from the eastern coastal regions with computing power from inland areas.

China is particularly making national efforts to develop AI foundation models, with local governments playing a significant role in this process. As exemplified by Peng Cheng Lab (PCL), Guangdong and Shenzhen are collaborating to enhance the synergy between data, computing power, and algorithms, which are key elements of AI competition. In September 2023, PCL announced Peng Cheng Mind (鹏城脑海), a Large Language Model (LLM), in collaboration with Huawei. This model uses 200 billion parameters and was pre-trained using computing power independently developed by China (Ding and Xiao 2023).

Leveraging areas where US containment is relatively weak is also a characteristic of China's response strategy. As clearly demonstrated by the export control measures on semiconductors from Nvidia and AMD in October 2023, US hardware export controls on China are expected to continue to strengthen in the future. While China has limitations in fundamentally resolving US export controls in the short term, it is responding through inventory expansion and imports via alternative channels (Wang 2024). In addition, China is focusing on developing open-source software and pursuing new innovation systems through cooperation with countries other than the US. The Chinese government's 60% financial support for joint research with Europe exemplifies this strategy.

While China currently employs a defensive strategy to counter US containment measures, it may seek a transition to an offensive and proactive strategy in the medium to long term. Considering China's rapid expansion of its manufacturing base for general-purpose semiconductors, it plans to first secure self-reliance in general-purpose semiconductors and then pursue a strategy where the US and its allies and partners become dependent on Chinese general-purpose semiconductors. This strategy aims to turn general-purpose semiconductors into a choke point for China.

China is also pursuing a strategy to surpass the US by enhancing its competitiveness in new semiconductor materials. This is a "corner-overtaking" (弯道超车) strategy, aiming to achieve comparable competitiveness in the development of new materials, as it is difficult to surpass the US's dominant position in silicon-based semiconductors. The market for Silicon Carbide (SiC) and Gallium Nitride (GaN) used in electric vehicles is expected to more than triple by 2026, and China's massive investment of $10.9 billion across 25 projects as of 2020 clearly illustrates this strategy (Lapedus 2021).

IV. US-China Competition and Global AI Governance

The fact that the US and China are pursuing conflicting AI paradigms not only foreshadows more intense US-China AI competition but also increases the likelihood of fragmented global AI governance. In particular, the trend of the US and China differentiating their ecosystems centered on core AI technologies is expected to strengthen. This trend is already detectable in AI talent mobility patterns. Fragmentation is also occurring in AI research. AI has been a field with active international joint research, but joint research between the US and China is decreasing. Furthermore, AI talent is increasingly moving within the US and China blocs, while mobility between blocs is weakening.

While the US absorbs a disproportionately large number of top-tier AI talents, the tendency for top AI talents to remain in their home countries is also increasing. Specifically, the mobility of top AI researchers has significantly decreased, as evidenced by the decline in the proportion of foreign AI researchers working in other countries from 55% in 2019 to 42% in 2022. This indicates the intensification of US-China AI competition and the growing national competition in AI.

The possibility of fragmented global AI governance is already apparent in the fact that major countries worldwide have announced their respective national AI strategies. As shown in <Figure 3>, major countries globally have uniformly announced national AI strategies. This reflects the recognition by countries worldwide of the significant impact AI will have on national competitiveness and security in the future. Simultaneously, individual national AI strategies are likely to make the establishment of global AI governance more challenging, as harmonizing AI strategies formulated based on national interests is not practically feasible.

<Figure 3> Countries Announcing AI Strategies

Source: Cohen, Jared, and George Lee. 2023. “The Generative World Order: AI, Geopolitics, and Power.” Goldman Sachs. December 14.

Given the significant geopolitical risks in the four key elements constituting AI competitiveness—computing power, talent, data, and infrastructure—the prospects for establishing global AI governance are not bright. Among these, geopolitical risks are assessed to be relatively higher in computing power and talent (Cohen and Lee 2023). Computing power is not only critically important for developing AI models, including AI semiconductors and data centers, but it is also a prime area of high geopolitical risk as it serves as a major means for the US to contain China in cooperation with countries like South Korea, Japan, Taiwan, and the Netherlands.

Talent is a core element constituting AI competitiveness, with the US, China, Europe, and India vying for dominance. As US-China competition intensifies, efforts to attract talent and prevent its outflow are expected to strengthen simultaneously. The data domain offers an advantage to countries with large populations, such as China and India; however, the US leads in applications required for utilizing and commercializing this data. While geopolitical risks are moderate, they are likely to increase in the future due to the intensifying US-China competition and the growing trend of data localization.

<Table 2> AI Competition and Geopolitical Risks

Source: Lazard. 2023. “The Geopolitics of Artificial Intelligence. October 17.

However, the possibility of establishing global AI governance is not entirely absent. The gap in AI capabilities is already widening among countries. While the US and China lead the competition, Canada, France, Israel, the UK, Singapore, South Korea, and India are assessed as countries that have independently built substantial AI capabilities (Cohen and Lee 2023). As these countries are exploring potential linkages with the US and Chinese AI ecosystems, efforts to prevent further separation of the US and Chinese AI ecosystems and to establish global AI governance at a limited level may be strengthened.

V. AI and the Global South

The effects of AI are already manifesting differently across countries, and this gap is expected to widen further. While some developed countries are reaping the benefits of AI, a growing number of developing countries are likely to fall behind in the AI race. In this context, China is likely to leverage cooperation with the Global South as a key means to expand the influence of its AI models. As shown in <Figure 4>, despite strong US containment, China has already exported Huawei equipment to numerous countries in Central Asia, Southeast Asia, Africa, and Latin America. This is a result of the Chinese government's support for companies entering countries participating in the Digital Silk Road (DSR).

China's centralized AI strategy is attractive to developing countries with authoritarian tendencies. These countries have a practical need to utilize AI not only for its commercial and economic benefits but also as a tool for regime stability and social control. The financial support and various incentives provided by China also contribute to developing countries' open attitude towards adopting Chinese AI models. China is likely to pursue a strategy of countering the establishment of US-led AI governance and proposing alternative governance by expanding and strengthening AI cooperation with the Global South. The Global South, in turn, may strengthen cooperation with China to escape AI blind spots and assert data sovereignty, thereby complicating the process of establishing US-led AI governance.

<Figure 4> China's Digital Infrastructure Participation Status

Source: Hillman, Jonathan, and Maesa McCalpin. 2019. “Watching Huawei’s ‘Safe Cities.’” CSIS.

The US-China AI competition is unfolding at different levels—science and technology, industry and economy, and infrastructure—and the nature of competition is expected to vary across these levels. Further breakdown reveals categories such as algorithms, computing power, commercial apps and services, and infrastructure. When all three levels are combined, the global AI order is likely to experience fragmentation centered around the US and China in the short term, and bloc-ization in the medium to long term. The bloc-ization of AI could lead to one of two scenarios: either the US-China competition escalates to an extreme, or conversely, a global governance order is established to mitigate the negative consequences of bloc-ization. In this process, an opportunity window may open for AI mid-tier powers to play a role in buffering the US-China AI competition (Effective Advocate 2024).

V. Future Outlook: From Technological Competition to Ecosystem Competition

The US and China may engage in even fiercer competition to secure superiority in core technologies, the most fundamental element for achieving AI leadership. Trends supporting this outlook are already emerging. Before the full-scale strategic competition between the US and China began, international cooperation, including joint research, was active. However, joint research between the US and Chinese scientific and technological communities is gradually decreasing, while joint research within respective blocs is intensifying and expanding. The rapid decline in AI talent mobility between the US and China, while mobility within blocs is increasing, also serves as an indirect indicator of the pursuit of core technological superiority by the US and China.

First, in addition to securing superiority in core technologies, the US and China will enter into competition to form AI ecosystems for the industrial or economic application of AI technologies. As sustainable revenue models for AI services have not yet been established, mergers and acquisitions centered around US and Chinese AI companies are expected to occur for a considerable period at the industrial level. During this process, the differences in the paradigms of the US and China pursuing AI ecosystem competition will become even clearer. In the US, the trend of innovation drivers shifting from universities and research institutes to corporations will strengthen in the process of industrializing and commercializing AI, and the link between research and development will become closer. Furthermore, as AI competition at the industrial level takes on the aspect of a "money game," competition among nations for resource mobilization will intensify. In this process, competition between China's model, where the state invests funds in the private sector, and the US model, which mobilizes funds from Big Tech and private investment, will become fiercer.

Second, the US and China are likely to pursue AI competition at the national strategy level for survival, moving beyond technological and industrial dimensions. In areas of advanced technology competition and industries that can affect national security, decoupling will be pursued, and industrial competition to preempt the AI market will intensify. However, in this process, the US is likely to leverage alliances and partners in the Indo-Pacific region and Europe, while China will likely utilize participants in the Belt and Road Initiative and the Global South as means for AI competition. Based on this, the US and China will likely engage in competition not only at the industrial level but also for the establishment of global AI governance favorable to their respective nations. The US will prepare for competition in AI norms and rules with China by minimizing disagreements on AI regulation with the EU. China will offer various incentives to the Global South to spread its AI models and attempt to export a more upgraded authoritarian AI model.

However, as indicated by the announcement of national AI strategies by major countries worldwide, national competition within the US and Chinese blocs may also accelerate. If some countries with AI capabilities enter the industrial-level competition, the fragmentation of the AI ecosystem could be promoted. To prevent this possibility from materializing, the role of AI mid-tier powers is required. ■

References

Bomey, Nathan. 2024. “Charted: U.S. is the private sector AI leader.” Axios. July 9. https://www.axios.com/...sector.

Cohen, Jared, and George Lee. “The Generative World Order: AI, Geopolitics, and Power.” Goldman Sachs. December 14. https://www.goldmansachs.com/...power.

Ding, Jeffrey, and Jenny W. Xiao. 2023. “Recent Trends in China’s Large Language Model Landscape.” Centre for the Governance of AI. April. https://cdn.governance.ai/...LLMs.pdf.

EET China. 2023. “China is Now Deciphering at Least 21 of the 35 Stalled Key Technologies!” (China is now deciphering at least 21 of the 35 stalled key technologies!).” April 9. https://www.eet-china.com/mp/a209546.html.

EffectiveAdvocate. 2024. “Middle Powers in AI Governance: Potential paths to impact and related questions.” Effective Altruism Forum. March 16. https://forum.effectivealtruism.org/...and.

Hillman, Jonathan E. and Maesea McCalpin. 2019. “Watching Huawei’s ‘Safe Cities.’” Center for Strategic & International Studies (CSIS). November 4. https://www.csis.org/...cities.

Lapedus, Mark. 2021. “China Accelerates Foundry, Power Semi Efforts.” Semiconductor Engineering. November 22. https://semiengineering.com/...efforts/.

Lazard. 2023. “Geopolitics of Artificial Intelligence.” Geopolitical Advisory Research Brief. October. https://www.lazard.com/...intelligence/.

MacroPolo. 2020. “The Global AI Talent Tracker 2.0 – 2023 Update.” https://macropolo.org/...tracker/.

The White House. 2022. “Remarks by National Security Advisor Jake Sullivan at the Special Competitive Studies Project Global Emerging Technologies Summit.” Briefing Room. September 16. https://www.whitehouse.gov/...summit/.

Tortoise Media. 2023. “The Global AI Index.” https://www.tortoisemedia.com/...#rankings.

Wang, Che-jen. 2024. “4 Ways China Gets Around US AI Chip Restrictions.” The Diplomat. June 28. https://thediplomat.com/...restrictions/.

Weinstein, Emily. 2021. “Don’t Underestimate China’s Military-Civil Fusion Efforts.” Foreign Policy. February 5. https://foreignpolicy.com/...efforts/.

■ Lee Seung-ju, Director of EAI Trade, Technology, and Transformation Center; Professor of Political Science and International Relations, Chung-Ang University.

■ Managed and Edited by:Park Ji-soo, EAI Researcher

    Inquiries and Editing: 02 2277 1683 (ext. 208) | jspark@eai.or.kr