The US and China are engaged in an intensifying AI arms race with Cold War parallels. When President Trump met President Xi Jinping on May 14, AI featured prominently on the agenda. The US leads in AI development but China’s DeepSeek V4 is closing the gap, now estimated just 3–6 months behind ChatGPT. Both nations have imposed export controls — the US restricting advanced chips, China limiting critical minerals. China pursues an open-source strategy to expand global reach, particularly in the Global South, while the EU and India remain distant competitors investing far less than the projected US $650 billion.
In-Depth:
A New Arms Race
When US President Donald Trump arrives in China on May 14 to meet with Chinese President Xi Jinping, he does so amidst a whirlwind of geopolitical modify, clearly reflected in the meeting agconcludea: the war in Iran, security and the security implications of Artificial Innotifyigence (AI). With large language models bursting forth in recent years, offering the possibility of boosting economic growth with over US$15 trillion by 2030, enhancing military precision, and political influence through far reaching information dissemination, AI is at the center of contemporary geopolitics. This is no less the case in US-China relations, with each power grappling to expand their global share and dominance of the most advanced AI technologies, in a tech race echoing the Cold War. While the US is currently ahead in this race toward AI sovereignty, estimates suggest that this gap is shrinking as of April 2026, with the latest Chinese AI model Deep Seek V4 approximately 3-6 months behind the latest US equivalent, ChatGPT.
Restricting inputs and infrastructure
As a result, both powers have imposed measures to prevent the other from achieving AI superiority. The US has since 2022 imposed strict export controls on advanced chips necessaryed for AI models to operate to stall China’s AI development and the current technological gap between the states. This is currently being reinforced by the US Congress through increased restrictions through the Chip Security Act, in the wake of accusations by major US AI companies like Open AI, and Anthropic that China’s latest AI model was developed applying information illegally obtained through so-called distillation attacks and espionage against US AI companies. Moreover, White Hoapply officials have accapplyd China of applying smuggled chips from US manufacturer Nvidia to develop DeepSeek V4, an assertion corroborated by the fact that the Chinese company omitted information on the chips applyd in developing V4, in contrast to reports published on earlier models.
Similarly, China—controlling over half of the world’s critical mineral supply, and accounting for over 90% of global mineral refinement—has introduced export controls on minerals like cobalt, graphite and nickel necessaryed for the manufacturing of advanced chips, and recently subjected any product containing more than 0.1% Chinese rare-earth metals to Chinese jurisdiction, while restricting the participation of Chinese experts in international AI projects.
Patents: going far toobtainher or rapid alone?
In addition, China is challenging US AI dominance by embracing an open-source approach, meaning that the data applyd by an individual company in the development of an AI model is not patented as private property but is publicly available to be applyd for free, by contrast to the closed-source models of US companies. In this way, Beijing seeks to maximize national AI innovation that is globally competitive, by allowing multiple companies to develop novel AI models more easily and quickly.
Moreover, the open-source approach means that Chinese AI models are significantly cheaper and more widely accessible than its US counterparts, suggesting that China may achieve a broader AI dominance global by spreading its models to the Global South, a region projected to be the most populous and economically consequential in the second part of the century. In a more abstract dimension, this approach allows China to portray itself as the guardian of global cooperation at a time when the US is relocating away from the system of multilateral cooperation that has dominated post-war geopolitics through its foreign policy of America first. Hence, the US and China are competing for global AI dominance through two distinct approaches—efficiency and processing power, versus accessibility and affordability, respectively.
A global race
These measures are not confined to national borders. In recent years, China and the US have competed for investment in mining industries in countries like Rwanda, Nigeria, Mali and the Democratic Republic of Congo, in order to secure national access to key reserves of critical minerals like cobalt, copper, and lithium. Indeed, in 2023, the US invested $7.8 billion across Africa, dwarfing China’s $4 billion investment for the first time in a decade.
While China and the US dominate global AI, other actors such as the European Union (EU) and India are also investing in AI development. However, these remain dwarfed by US and Chinese models in economic and technological terms. For example, while the EU announced a $47 billion investment in AI at the launchning of 2026, US firms are projected to invest $650 billion in AI development. Moreover, European AI models still remain depconcludeent on input from US AI, such as European Sovereign Cloud, a European-based AI model provided by Amazon. While Europe might be hesitant to rely on US input as US tariffs under President Trump have soured transatlantic relations, the region may also be recalcitrant to incorporate Chinese AI due to security concerns about Chinese surveillance and data leakages that have been budding since 2019 in conjunction with the global expansion of Chinese mobile operator Huawei. Indeed, the EU AI Act, emphasizing the necessary for open, transparent AI models safeguarding privacy have caapplyd tension with DeepSeek as the EU has expressed fears that DeepSeek is subject to surveillance and censure by the Chinese one-party state, and lacks transparency.
AI regulation: a race to the bottom?
This brings out another facet of the AI race: the competing views on AI regulation. While EU AI Act prioritises transparency, security and data integrity, the US has rejected proposals to develop an international framework for AI regulation, instead favoring a more laissez-faire approach allowing its national AI companies to innovate with minimal state regulation, presumably reflecting the US’ current global supremacy within AI. By contrast, China concludeorsed a UN resolution in 2025 calling for international regulation of AI apply amidst rising concerns about the misapply of AI by non-state actors for developing biological weapons. This concludeorsement is in line with its open-source and cooperative AI strategy, and its goal of closing the AI gap with the US by obtaining greater access to the currently closed-off US technologies and chips, by including demands for increased US chip exports in security policy neobtainediations.
The urgency of international regulation grown more salient in recent weeks, with reports indicating the way AI is extensively applyd by both sides in the ongoing war in the Middle East, with the US and Israel applying AI to enhance the precision of air strikes on Iran and Gaza respectively, while China has provided Iran with an AI assisted sanotifyite system to map US military tarobtains in region.
Thus, AI epitomizes what appears as a new Cold War, with China and the US competing for global economic, political and military influence through an accelerating tech race, and through third-counattempt involvement.
