Intensifying US-China Competition in AI and Semiconductor Technology: Selective Decoupling Evident in China's Data Control and Western Firms' Withdrawal

Executive Summary

Beijing's strong reaction to Meta's attempted acquisition of Chinese AI startup Manus has led to strengthened controls by the Chinese government on the entry and exit of data and AI technologies. This, coupled with Nokia's withdrawal from the Chinese market and the accelerated delisting of Windows from government institutions, signals a phase where decoupling in technology sectors is becoming institutionalized. This trend is most likely to manifest as a selective separation, confined to areas directly linked to national security such as AI, semiconductors, and telecommunications equipment, rather than a complete severance. Korean companies and the government should adopt a dual-track approach: proactive risk management for strategically vital national technologies and continued monitoring with maintenance of the existing approach for general commercial sectors, rather than binary choices like complete withdrawal or maintaining the status quo. Regarding pressure to choose sides under the US's Pax Silica initiative, maintaining a principle of sector-specific responses rather than prematurely fixing a stance would be more beneficial.

I. Issue Analysis

Intensifying US-China Competition in AI and Semiconductor Technology: Decoupling Amidst China's Tightening Data Controls and Nokia's Withdrawal

1. Background and Progression of the Issue

The direct trigger for Beijing's recent strengthening of entry and exit controls was Meta's attempt to acquire the Chinese AI startup Manus. Meta sought to acquire Manus, a developer of agentic AI, in December 2025 [9]. This transaction provoked a strong backlash from the Chinese government. Beijing demanded the dissolution of the deal to maintain control over Manus within its jurisdiction [9]. Consequently, Tencent acquired Meta's stake, becoming the largest shareholder in Manus [9].

Nikkei Asia highlighted this incident as the catalyst for a policy shift by Chinese authorities. It reported, "China's government has bolstered entry and exit controls to prevent leaks of data, artificial intelligence, and other advanced technologies with rules slated to take effect on Sept. 15, aiming to shore up economic security" [1]. The new regulations are set to be implemented starting September 15 [1]. The core of these regulations is to preemptively block channels through which data and AI technologies could be leaked abroad via the movement of personnel and capital [1].

This is not an isolated measure. Following the shock of DeepSeek's model release, China has reportedly elevated the deployment of open-weight models to a national strategy [8]. In conditions where access to advanced semiconductors is blocked by US export controls, China is seen as shifting its competitive front to expanding the developer ecosystem at the grassroots level [8]. The Manus incident can be viewed as an event that occurred in the blind spot of this national strategy. The sense of crisis that the nation's AI assets could fall under foreign capital control led to Beijing's swift intervention.

Nokia's withdrawal from China proceeded on a separate track but aligns with the same trend. The Edge Malaysia, citing the South China Morning Post, reported that Nokia plans to phase out most of its operations in China within the year and significantly reduce its workforce [17]. Nokia explained the background as "we are adjusting our operational footprint in China to address this reality," citing years of continuous contraction in its Chinese business [17]. Digitimes Asia specifically reported the closure of the Hangzhou R&D center and a workforce reduction of 1,600 employees, analyzing that shrinking market access, geopolitical friction, and global restructuring had cumulatively contributed to this decision [18].

2. Current Situation

China's data and technology controls and the market exit of Western firms are two parallel developments. The accelerated delisting of Microsoft Windows from Chinese government institutions is also an extension of this trend [19]. This appears to be the accelerated implementation of an existing policy to reduce reliance on foreign software in critical national infrastructure.

In the realm of AI model competition, DeepSeek has officially launched V4 Pro just four months after its preview. It features a Mixture-of-Experts (MoE) structure with 1.6 trillion parameters, activating approximately 49 billion parameters per token [7]. It supports a 1 million token context window and an output of up to 384,000 tokens [7]. The Global Times reported that DeepSeek has introduced a tiered pricing system for peak and off-peak hours while increasing API fees. Analysts interpret this as a signal of transition towards "healthier commercialization" [11]. This appears to be an attempt to move beyond the free distribution phase and enter a stage of profitability.

Nikkei Asia reported that the release of this open-weight model has once again caused ripples in Washington's Silicon Valley and policy circles. It noted, "Washington and Silicon Valley debating more intensely than ever whether -- and how -- the U.S. should respond to the rise of such Chinese AI technology" [4]. The Brookings Institution expresses similar concerns, analyzing why Washington is worried about the competitive landscape between China's open-source models and US companies like Anthropic and OpenAI [6].

The United States appears to be expanding this competition from a bilateral technological contest to a bloc-based confrontation. According to multiple media reports citing Reuters, the US Department of State is preparing to demand that various countries choose sides in the AI competition. It warns that countries that simultaneously join China's competitive framework will be excluded from the US-led AI alliance [14][15][16]. Through the Pax Silica initiative launched in 2025, the US aims to secure supply chains for AI models, semiconductors, and critical minerals [15]. Reports indicate that some countries, including Kazakhstan, are already being forced to make choices between the US and Chinese frameworks [16].

Competition is also expanding in the semiconductor and hardware sectors. The South China Morning Post reported that the US Federal Communications Commission (FCC) is considering a ban on Chinese-made optical transceivers. This regulation targets key components used for high-volume data transmission in AI data centers and cloud networks [13]. The control is extending beyond semiconductor chips to encompass detailed components of AI infrastructure.

3. Key Actors and Positions

Chinese GovernmentIn this phase, the Chinese government prioritizes preventing the outflow of its AI assets abroad. The method of reorganizing shareholding structures through Tencent in the Manus incident demonstrates this [9]. The strengthening of entry and exit controls, effective September 15, also aims to preemptively block technology leaks through the movement of personnel and capital [1]. Concurrently, the government is accelerating the delisting of Windows from its institutions, working to reduce reliance on foreign technology in critical domestic infrastructure [19].

Chinese AI Companiesare positioned at the intersection of national strategy and commercial interests. Tencent, through the Manus acquisition, has complied with the national demand for maintaining domestic control while securing assets [9]. DeepSeek is moving away from its previous strategy of expanding the global developer ecosystem through free open-weight model distribution and is now attempting a transition to a revenue model by increasing API fees and introducing tiered pricing [11]. Subsequent models like Alibaba's Qwen are also participating in the industry-wide open strategy by maintaining an open-weight approach [8].

Nokiaserves as an example of a Western company that has decided to withdraw from the Chinese market. The company used the euphemism "alignment with global operations" rather than directly mentioning geopolitical factors [17]. However, Digitimes Asia's analysis points to shrinking market access and geopolitical friction as the substantive reasons for the withdrawal [18]. The closure of the Hangzhou R&D center is particularly symbolic, signifying not just a reduction in business operations but a decision to relocate R&D functions out of China [18].

Metais the party that succumbed to pressure from the Chinese government in this phase. It attempted to acquire Manus but was forced to transfer its stake to Tencent due to Beijing's demand for deal dissolution [9]. This serves as a case study illustrating the concrete political constraints faced by Western big tech companies in accessing Chinese AI startups.

US Governmentseeks to expand the AI competition from a bilateral technological contest to an issue of bloc selection. Reports that the State Department is preparing to require partner countries to choose between US and Chinese frameworks support this [14][15][16]. The FCC's review of regulations on Chinese optical transceivers is an attempt to extend control beyond finished semiconductor products to the component-level supply chain [13].

4. Key Issues

The first issue is the potential conflict between China's intensified data and AI controls and its efforts to attract foreign capital for domestic companies. As seen in the Manus case, the very channels through which foreign capital enters Chinese AI startups are now subject to national security review [1][9]. This could potentially shrink the global fundraising environment for Chinese AI startups.

The second issue concerns whether Western companies' withdrawal from China is a matter of individual corporate commercial judgment or a result of structural decoupling. While Nokia uses euphemisms like business alignment [17], the substantive action of closing its Hangzhou R&D center signifies a reduction in the very foundation of technological cooperation [18].

The third issue is the emergence of open-weight AI models as a new front in the US-China competition. The free distribution of DeepSeek V4 Pro directly pressures the business models of US AI companies [4][6][7]. Simultaneously, DeepSeek itself faces pressure to secure profitability and has entered a phase of adjusting its pricing structure [11].

The fourth issue is the extent to which the US-driven bloc pressure can be effectively implemented. Reports indicating that some countries, like Kazakhstan, are already at a crossroads [16] suggest that many countries that have pursued their interests between the US and Chinese frameworks may face policy dilemmas in the future. For third countries, including South Korea, the need to establish response principles by distinguishing between critical national infrastructure and private industry sectors becomes even more apparent in this context [8].

II. In-depth Issue Analysis

Intensifying US-China Competition in AI and Semiconductor Technology: Underlying Causes and Structural Context

1. Analysis of Underlying Causes

While the immediate trigger for this phase was ostensibly the single event of Meta's attempted acquisition of Manus, the speed and intensity of Beijing's response indicate it was not an accidental reaction. The Chinese government immediately demanded the dissolution of the deal after the acquisition announcement [9]. The time it took for Tencent to acquire Meta's stake and become the largest shareholder was not long [9]. This swiftness suggests that Beijing was already constantly vigilant about the possibility of AI assets being leaked abroad.

The underlying causes can be divided into two categories. One is the structural constraint created by US semiconductor export controls targeting China. With access to advanced semiconductors blocked, China adopted the expansion of its developer ecosystem as an alternative national strategy [8]. The success of this strategy hinges on whether domestic AI companies remain under national control. Events like the potential acquisition of promising agentic AI companies like Manus by foreign capital are perceived as threats that undermine the foundation of this strategy.

The other cause is the Chinese government's long-standing sensitivity regarding data sovereignty. The revision of entry and exit regulations reflects an intention to fundamentally block the leakage of data, AI, and advanced technologies through the movement of personnel and capital [1]. Nikkei Asia characterized this as a measure "aiming to shore up economic security" [1]. While the rationale of preventing technology leaks is not new, this measure differs from previous ones in that it concretized regulations and set a specific implementation date following a particular incident.

Nokia's withdrawal, however, stems from a different set of underlying causes. Digitimes Asia analyzed that shrinking market access, geopolitical friction, and Nokia's global restructuring had cumulatively contributed to the decision [18]. Nokia itself cited the years of continuous contraction in its Chinese business as a background factor [17]. This can be seen as a cumulative result of the increasing market share of domestic companies like Huawei and ZTE in the telecommunications equipment market and the exclusion of foreign companies from government procurement, rather than direct pressure from the Chinese government. The accelerated delisting of Windows from government institutions also falls within this context [19]. It appears to be an accelerated implementation of an existing policy to reduce reliance on foreign software and hardware.

2. Structural Context

At the political level, this issue is directly linked to the Xi Jinping leadership's drive for technological self-reliance. As long as US pressure regarding semiconductors and AI continues, the Chinese leadership has few options other than to strengthen national control over critical technology assets. The Manus incident exemplifies how this drive can clash with the autonomous M&A activities of private companies. Beijing prioritized technological sovereignty over market logic.

At the economic level, the exit of foreign companies from the Chinese market is becoming structurally entrenched. Nokia's withdrawal is not limited to the telecommunications equipment sector. As the exclusion of foreign products from government procurement expands, the profitability of Western companies reliant on these sectors will structurally deteriorate. Nokia's closure of its Hangzhou R&D center and the reduction of 1,600 jobs [18] indicate that the consequences have reached the stage of actual employment and investment withdrawal. This represents a different pressure pathway than for companies directly exposed to the consumer market, such as Apple or Tesla. In the B2B and B2G-oriented telecommunications equipment market, changes in government policy are immediately reflected in sales.

At the security level, there is a clear trend of data and AI models being treated as new strategic assets. The US Department of State is reportedly considering demanding that partner countries choose sides in the AI competition [14][15][16]. Washington's Pax Silica initiative, launched in 2025, aims to secure supply chains for AI models, semiconductors, and critical minerals under US leadership [14]. Reports indicate that several countries, including Kazakhstan, are being forced to choose between the US and Chinese frameworks [16]. This demonstrates that technological competition is escalating from the individual corporate level to inter-state bloc formation. The FCC's review of regulations on Chinese optical transceivers [13] suggests that this competition is targeting not only semiconductor chips but also the underlying infrastructure of AI data centers.

3. Historical Precedents and Comparison with Similar Cases

China's current strengthening of entry and exit controls is consistent with the enactment of the Anti-Foreign Sanctions Law in 2020 and the implementation of the Data Security Law in 2021. In those instances as well, specific events triggered legislation. Examples include the controversy surrounding the suspension of Alibaba's Ant Group IPO and data security concerns related to Didi Chuxing's US listing. The pattern of regulations being rapidly refined after an incident and having a clear implementation date is recurring. In this case, the short period between the Manus incident and the regulation's implementation also confirms a similar response pattern [1].

Nokia's withdrawal is an extension of the trend of Western telecommunications equipment companies scaling back their operations in China since the late 2010s. Ericsson has also faced similar pressures. The pattern of foreign telecommunications equipment companies facing shrinking market share due to the growing dominance of domestic players like Huawei and ZTE has been observed multiple times. However, the current withdrawal is more pronounced in its scale and speed compared to previous cases. The inclusion of R&D center closures signifies not just a reduction in business operations but the cessation of R&D functions within China [18].

In the context of US-China AI competition, the developments following the release of DeepSeek's R1 serve as a benchmark. Following DeepSeek's success in creating cracks in the US-led narrative with its low-cost, high-efficiency model, subsequent models from Alibaba's Qwen, Zhipu AI, and Moonshot AI have been released using the open-weight approach [8]. This has led to the analysis that it has become an industry-wide approach rather than an individual company's choice [8]. The current V4 Pro release and API price increase [7][11] indicate that this approach is entering a phase where its commercial sustainability is being tested in the market. This follows a trajectory similar to the open-source software community's search for commercialization models in the 2000s.

4. Key Variables Shaping the Issue's Development

The first variable is the actual enforcement intensity of the entry and exit regulations taking effect on September 15. Regardless of the regulations being codified, the key question is the extent to which they will impose practical restrictions on foreign capital investment in Chinese AI startups and personnel exchanges. If the regulations are applied too strictly, there is a risk of stifling the inflow of foreign capital and talent that the Chinese AI industry needs.

The second variable is whether the withdrawal of Western companies from the Chinese market will spread beyond telecommunications equipment to other sectors. The key observation point will be whether Nokia's case remains an isolated corporate restructuring or signals a broader exclusion from government procurement extending to other fields.

The third variable is whether the US-driven bloc pressure will lead to the actual acquisition of participating countries. If the Pax Silica initiative gains substantial support from numerous countries, the bifurcation of the AI supply chain will accelerate [14][15]. Conversely, if, as in the case of Kazakhstan, many countries defer clear choices between the US and Chinese frameworks [16], bloc formation may remain a rhetorical slogan, and a mixed supply chain could persist for the time being.

The fourth variable is the commercial sustainability of Chinese open-weight AI models. Depending on whether DeepSeek's API price increase [11] is a normal adjustment for profitability or a signal of the financial limitations of the open-weight strategy, the very front of China's AI competition may be readjusted.