The AI cold war that could reshape global power

.

By early 2024, Beijing was deeply unsettled by how far American companies had pulled ahead in generative AI. OpenAI, Google and a cluster of Silicon Valley start-ups were driving global excitement, while Chinese firms were so far behind that many relied on Meta’s open-source Llama. On top of that, US restrictions on exporting advanced chips were threatening to handicap China’s ability to catch up. Under pressure, Beijing began calling in tech executives, easing regulations, injecting capital and accelerating the rollout of computing power across the country, the Wall Street Journal reported.

China’s breakthrough and the shock it sent through Silicon Valley

The mood shifted in early 2025, when a young Chinese company, DeepSeek, released a model that stunned industry insiders with its speed and reasoning ability. Premier Li Qiang reportedly told officials that China now had a model it could take pride in. The breakthrough triggered a fresh burst of state support and new urgency across China’s tech sector. In the United States, DeepSeek’s rise jolted Silicon Valley and sharpened warnings that Chinese progress was quickening despite chip constraints.

The geopolitical stakes and the Cold War echoes

The competition now resembles a modern version of the Cold War’s technological races. Both countries see AI as a transformative capability that could reshape the global economy and military balance. In Washington, fears revolve around what officials call China’s “authoritarian AI,” a system that could export censorship and influence. In Beijing, there is a conviction that falling behind in AI would give the US leverage to block China’s ambitions as a global power. Both sides increasingly believe that dominance in AI means influence over how billions of people work, learn and communicate.

China’s push for national-scale AI development

Beijing’s strategy relies on scale. A nationwide campaign aims to build huge clusters of data centres in regions with abundant renewable energy and to link them into a shared compute network that supporters call a national cloud. Local governments are providing cheap access to computing power, setting up public data sets and subsidising researchers. China has also eased safety rules to allow faster deployment, reversing earlier caution that had required companies to complete lengthy content reviews before models could be released.

America’s current edge and its vulnerabilities

The United States still produces the world’s most powerful frontier models and dominates the design of advanced chips that make training them possible. Venture capital is pouring hundreds of billions into AI start-ups, giving American firms an advantage in both talent and experimentation. Yet there are vulnerabilities. China has more engineers, lower operational costs and a coordinated state-led model that can mobilise resources at extraordinary speed. If advances in chips begin to plateau, or if frontiers shift toward new architectures, the American lead could narrow.

How China is trying to solve its chip problem

Unable to buy the most advanced AI chips, China is improvising. Companies such as Huawei have been working with thousands of local suppliers to build their own semiconductor ecosystem. One strategy bundles hundreds of thousands of lower-grade domestic chips into large compute systems that try to match the performance of top-end American hardware. It uses more power and is less efficient, but it offers a potential path to capability that Beijing sees as vital.

Why this contest could define the next decade

For both the US and China, AI is not only a question of commercial advantage but a test of national resilience. It will affect defence planning, cyber security, biotechnology, information control and the future of labour markets. As each side accelerates, cooperation becomes harder and the risk of espionage, hacking and intellectual property theft rises. Leaders on both sides say the outcome will shape scientific progress and global influence for a generation. The speed of the race, and the depth of mistrust surrounding it, suggest the geopolitical world that emerges from it may look very different from the one that existed before the AI boom began.