China-US Competition to Lead the AI Race Mirrors the Nuclear Arms Race
CNN reported that China is moving to consolidate its position as a leading global power in artificial intelligence (AI), after the government unveiled a comprehensive strategy aimed at placing the country at the forefront of the sector by the end of the decade. This effort leverages massive funding and resources, as well as the nature of its political system, which enables coordination between state institutions and the tech sector.
The website added that while the Chinese Communist Party may exaggerate some of its goals, Chinese leadership views AI as an existential issue for the future of its rule, prompting it to mobilize its capabilities to address this challenge. At the same time, China has already begun narrowing the gap with the United States in this field.
The US AI industry, meanwhile, appears more preoccupied with the risks posed by the technology itself. The prospect of developing systems that could escape developers’ control and ultimately lead to catastrophic consequences for humanity is raising growing concerns within Silicon Valley.
The potential for AI to cause widespread destruction, and the likelihood of reaching that stage, remains a subject of intense debate without definitive scientific resolution. Nevertheless, the United States and China are engaged in a technological and strategic race that, in some respects, resembles the nuclear arms race between the two superpowers during the Cold War, when the danger of a shared threat forced both sides to the negotiating table to establish rules for risk reduction.
The US government has taken steps to maintain its technological edge, including imposing restrictions on China’s access to the latest AI chips, which has created significant obstacles for Chinese companies seeking to develop more advanced models.
In response, Chinese companies have developed strategies to circumvent these restrictions by improving software efficiency and reducing their reliance on advanced computing capabilities.
In February 2025, the Chinese model “DeepSeek” lagged behind OpenAI’s best models by only 0.4% in standardized tests, a progress considered significant compared to previous Chinese models. Since then, Chinese and American models have become closer in performance, according to the 2026 AI Index Report issued by Stanford University, which noted that the performance gap between the two countries “has already closed.”
The United States still retains important advantages in producing advanced models, private investment, and data center infrastructure, according to Andrew Chen, a professor of AI law at the University of North Carolina, who emphasized that determining which country is “leading” in this race depends largely on the metrics used for measurement.
China has not abandoned its ambitions, revealing in 2025 the “New Generation AI Development Plan,” which aims to enhance the country’s position in transformative technologies and leverage the technology sector and government institutions under the Party’s supervision to develop infrastructure and train large AI models.
China’s strategy is not without drawbacks; researchers at the University of Virginia argue that the nature of the Chinese system opens the door to political cronyism and increases the likelihood of bias in AI system outputs.
Meanwhile, the US AI industry continues to rely heavily on voluntary commitments regarding self-regulation, as sector leaders meet with political officials to discuss ways to mitigate the risks of the technology.
US President Donald Trump criticized such efforts, arguing that excessive regulation could amount to capitulation to fears and misconceptions about artificial intelligence, and could also give China an edge in the technological race.
The construction of data centers in the United States is falling short of the ambitious goals set by companies to expand their computing capacity, amid shortages of labor, construction materials, and advanced chips, as well as difficulties in obtaining permits, rising interest rates, and local community opposition to data center projects.
The core problem is that no one actually knows how much computing power AI needs to reach a level that poses an existential threat to humanity.
