From Cooperation to Competition: Is America Making China Stronger?

The once deep and successful scientific and academic collaboration between the United States and China is undergoing a sharp separation. While partnerships like that between Microsoft Research Asia (MSRA) and Tsinghua University once produced groundbreaking research like the ResNet paper, the dynamic has now changed dramatically. Under political pressure, major American tech companies are closing or downsizing their research labs in China, while US lawmakers are urging universities to end joint programs with Chinese entities like the Chinese Scholarship Council. This decoupling aims to curb technology transfer, but it is leading to a progressive fragmentation of the global research ecosystem

Relations between the United States and China, once characterized by solid cooperation in science and technology, are experiencing a profound decoupling. What was once a symbiotic collaboration between the two countries' academia and industry is now giving way to strategic competition, which also extends to student exchanges. The signs of this friction are tangible: Science and Technology Cooperation Agreement (STA), a cornerstone of the partnership for decades, has been the subject of tense negotiations and short-term renewals, reflecting eroded political trust. Added to this are the closures of research laboratories of American giants in China and legislative proposals in the United States aimed at limiting Chinese students' access to STEM programs, marking a sharp departure from a model of openness that once fostered the growth of both countries.

Until recently, the interconnection between the two countries' research ecosystems was profound. The architecture for this partnership was established on January 31, 1979, with the signing of the STA by Jimmy Carter and Deng Xiaoping. US universities have educated over 90,000 Chinese PhD students, with approximately 70% of them specializing in STEM fields. An estimated 80 percent of these talents remained in the United States, contributing significantly to American human capital and research capabilities. This golden era of cooperation has been exemplified by institutions such as Microsoft Research Asia (MSRA) in Beijing and Tsinghua University, both products of this partnership. According to the Journal of Quantitative Science Studies published by the MIT Press, China and the United States have become each other's major partners in high-impact research publications, clear evidence of the depth of this collaboration.

The depth of this collaboration has produced research of global importance. An example is the paper Deep Residual Learning for Image Recognition (commonly known as ResNet), proposed by He Kaiming during his research at MSRA. This paper is considered the most cited scientific paper of the 21st century, and its "residual connections" have become ubiquitous in modern deep learning models, from Transformers to AlphaGo Zero and nearly all current generative AI models. He Kaiming himself, before working at MSRA, earned a bachelor's degree from Tsinghua University and interned at Microsoft's research lab in Beijing, demonstrating the deep interconnectedness between the two institutions. The connection between MSRA and the Chinese tech ecosystem is further highlighted by the fact that another author of the ResNet paper, Xiangyu Zhang, now works at Megvii, a major Chinese AI company that has been blacklisted by the US government.

Today, however, this partnership has entered a phase of profound uncertainty, slipping from a shared engine of progress to a battleground of strategic competition. The most obvious effect is a geopolitical paradox:In trying to contain Beijing's technological capabilities, the United States is unwittingly accelerating China's push toward self-sufficiency.

The New American Manual: A Multi-Level Containment

The United States' posture toward China has shifted from viewing the relationship as mutually beneficial to one of intense competition. Scientific and academic collaborations are now viewed through the lens of national security, with growing concerns about technology transfer and espionage. This shift has led Washington to implement a sophisticated, multi-layered approach, commonly referred to as the "new playbook" for containment.

The most tangible sign of this tension is the withdrawal or reorganization of U.S. companies' research activities in China. While Google closed its AI research lab , Amazon disbanded its research center in Shanghai due to “increasing political pressures from the United States” and “strategic adjustments in the context of U.S.-China tensions.” This closure occurred despite the lab having published over 100 research papers and created an open-source neural network framework that had generated approximately $1 billion in revenue for Amazon. Other major companies have also scaled back their operations: IBM has cut more than 1,000 R&D jobs in 2023, and Microsoft has offered hundreds of employees from its cloud and AI divisions the opportunity to relocate abroad in 2024.

Microsoft's approach to the MSRA is a case study in "managed competition." The company admitted implementing new "guardrails" to prevent China-based researchers from working on areas considered politically sensitive, such as facial recognition and quantum computing. This move is a direct response to criticism that academic collaboration, including co-production of AI research with the military-run National University of Defense Technology (NUDT), has strengthened China's military and surveillance capabilities.

In addition to these corporate adjustments, the US government has used a variety of tools to implement technological containment.

• Export controls. Beginning in October 2022, the United States implemented highly specific export controls on advanced semiconductors. The restrictions target logic chips measuring 16 nm or less, DRAM memory chips measuring 18 nm or less (half-pitch), and NAND flash chips with 128 or more layers. The goal is clear: to limit China's ability to design, produce, and access AI chips and related technology, particularly to prevent their use in military and surveillance applications. These controls also aim to limit human capital, forcing US citizens and green card holders to leave the Chinese semiconductor industry.

• Targeted sanctions.In 2019, Chinese AI company Megvii Technology Ltd. was inserted. Megvii was placed on a US trade blacklist for its alleged involvement in human rights abuses in the Xinjiang region. This designation, which limited its ability to purchase components from US suppliers, made Megvii's initial public offering (IPO) in Hong Kong a "hard sell" for many American investors. It's important to note, however, that there was significant confusion between two individuals with similar names. The US sanctions do not apply to Li Shipeng, one of the founders of the MSRA, but to General Li Shangfu, China's former Minister of Defense, for purchasing military equipment from Russia.

• Legislative pressure on universities.This restrictive approach also extends to the academic sector. In July, Chairman Moolenaar (R-Mich.) of the Special Committee on China envoy letters to seven universities (including Dartmouth College, Temple University and the University of Tennessee) urging them to end their joint programs with the Chinese Scholarship Council (CSC), the body of the Chinese Ministry of Education that provides support for international academic exchanges.The committee views the CSC as a Chinese Communist Party (CCP)-led initiative that leverages U.S. institutions for technology transfer to benefit China's military and scientific growth. According to reports, recipients of these scholarships are required to sign oaths of loyalty to the CCP and receive "ideological training" before traveling abroad, highlighting a clear concern about political influence. The action follows the relationship of the committee, "CCP on the Quad," released last September, which exposed how hundreds of millions of dollars in U.S. taxpayer-funded research has helped improve China's military capabilities.

• New economic agreements. In an unprecedented development, the United States has explored the monetization of trade policy. As a condition for resuming sales of specific AI chips to China, Nvidia and AMD agreed to pay 15% of their revenues to the US government. This agreement drew sharp criticism from lawmakers, who called it a "dangerous abuse of export controls" and a "creative tax scheme disguised as national security policy."

On top of all that,theTensions between Washington and Beijing continue to evolve, with recent announcements further defining the picture of the competition. 

On August 13, the White House confirmed that the Trump administration is finalizing an agreement to impose a 15% export tax on Nvidia and AMD chip sales to China. This agreement, in exchange for export licenses for AI chips specifically designed for the Chinese market, could be extended to other companies in the future. President Trump wants the companies to "pay us as a country" for the issuance of these licenses, although the legality of the plan is still a matter of debate among trade experts and lawyers.

In turn, the Commerce Department has announced actions to further strengthen export controls on AI chips. The new guidance aims to warn industry about the risks of using advanced Chinese chips (such as Huawei Ascend chips) and to alert U.S. industry to the possible consequences of allowing AI chips to be used for training Chinese models. At the same time, the trade truce between the United States and China, which was scheduled to expire, has been extended by 90 days, until November 10, 2025. This temporary delay is intended to avoid potential economic escalation and allow for further negotiations, a decision experts attribute in part to China's influence over critical minerals, such as rare earths.

China's Response: A Catalyst for Autonomy

Rather than stifling China's technological development, US measures have acted as a powerful catalyst for accelerating a pre-existing national self-sufficiency strategy. Initiatives such as "Made in China 2025" (MIC2025), launched in 2015, and the 2017 "Next Generation AI Development Plan" already aimed to transform China from a "world's factory" to a high-tech powerhouse by reducing dependence on foreign suppliers. The MIC2025 plan, in particular, was heavily inspired by the German "Industry 4.0" program and served as a "directional roadmap" for innovation, encouraging companies to reduce their dependence on foreign suppliers through government incentives, direct R&D funding, and policies favoring the acquisition of technology by Chinese firms.

Progress has been lopsided but significant. Although China has struggled to reach its goal of 70% self-sufficiency in semiconductors by 2025, it has nevertheless seen exponential growth in the sector, with the number of registered chip companies rising to 22,800 by 2020 and sales more than doubling over the same period. Although a gap in the most advanced chips remains due to a lack of "intangible skills" and equipment such as EUV photolithography machines, the Chinese industry has demonstrated its resilience.

In other sectors, success is unmistakable. China has become a world leader in electric vehicle production, exceeding 10 million units annually, and dominates over 75% of global lithium-ion battery production. China's achievements are not limited to the automotive industry; the country has reached impressive milestones in other high-tech areas, such as the Zuchongzhi 3.0 quantum computer, which operates trillions of times faster than classical supercomputers, and the domestically produced AG600 amphibious aircraft. Its high-speed rail network is the largest and most advanced in the world, with more than 48,000 km in service, and its AI model, which excels in "wide adoption" rather than just high-end innovation, is spreading its technology throughout the developing world through initiatives such as the Digital Silk Road.

Consequences and the bifurcated future

Economic analysis of a complete technology decoupling shows welfare losses for both countries. In a unilateral scenario in which the United States completely cuts off technology flows to China, Chinese welfare would decrease by 1.65 percent, while a trade ban alone would reduce welfare by only 0.014 percent. This finding from a study using a dynamic quantitative trade model underscores that restricting the flow of knowledge is far more damaging than restricting trade in goods, highlighting the strategic importance of human and intellectual capital in the current rivalry.

In this context, the rivalry between the United States and China is shaping up as a competition between two distinct innovation models. The US model, driven by private sector giants, excels at high-end innovation and the creation of cutting-edge AI models. However, this same competitive environment has pushed US companies to differentiate themselves and innovate in new areas, as a strategic response to Chinese competitive pressures. In contrast, the state-led Chinese model excels at broad adoption and the large-scale integration of AI into the economy and the state. The long-term risk for the United States is that, while maintaining leadership in cutting-edge innovation, it may lose the competition for market share and global influence to a more accessible and widely adopted Chinese ecosystem.

What's clear, ultimately, is that the dynamic between the two countries has created a profound paradox. US containment measures, while potentially slowing China's access to cutting-edge technologies, are fueling its determination to become a technologically self-sufficient entity. China, driven by fear of technological "strangulation," is building an indigenous ecosystem that could emerge as an even more formidable and resilient competitor in the long run.