US-China AI race accelerates as open models and semiconductor self-sufficiency reshape strategic dominance

Despite efforts to restrict China’s technological advancement, the US and China are engaged in a rapid race for AI, robotics, and semiconductor dominance, with open-source models and self-reliance tactics prompting a disruptive shift in global tech leadership.

Washington’s long campaign to slow China’s technological rise is increasingly producing the opposite effect, according to reporting from the source material. Instead of halting progress, export controls, investment curbs and procurement bans have pushed Chinese firms towards faster domestic innovation in artificial intelligence, robotics and semiconductors.

That shift became harder to ignore after DeepSeek showed that a Chinese laboratory could build a capable AI system without cutting-edge hardware. Moonshot AI then intensified the challenge in July with Kimi K3, which was presented as an open-weight model with 2.8 trillion parameters and positioned as competitive with leading systems from OpenAI and Anthropic. Because the model’s weights are openly available, developers can inspect, adapt and build on it more easily than with closed commercial systems, which makes it strategically significant for both adoption and competition.

The debate in the US has since split along both security and commercial lines. OpenAI and Anthropic argue that Chinese open models create risks that justify tighter restrictions. Nvidia, Microsoft, Meta, Palantir and a range of start-ups take the opposite view, saying open-weight models can accelerate the broader industry and that a ban would hurt American business more than it would help national security. That dispute now goes beyond geopolitics and has become a direct argument over market structure.

The security argument is not limited to software. Recent reporting from TechRadar described a Hugging Face cybersecurity incident involving a Chinese-developed GLM-5.2 model, underscoring how open systems can be exposed to tampering or hidden code. That case has added to the pressure in Washington as policymakers weigh whether open-weight AI should face new limits.

In robotics, China already appears to have the upper hand. According to figures cited by Kronen Zeitung, Chinese companies such as Unitree, AgiBot and UBTech control roughly 80% of the global humanoid robot market. About 16,000 such robots were sold in 2025, and analysts expect annual sales to reach about 100,000 units by 2027. These machines are no longer just demonstration products; Beijing sees them as future labour for factories, warehouses and logistics networks.

That is one reason the US Federal Communications Commission has moved to block imports of certain connected bipedal and quadrupedal robots from China. Washington fears they could be used for spying or remote control. Beijing has responded by accusing the US of trying to preserve “AI hegemony”, while Chinese officials continue to frame robotics as a strategic industrial priority.

Semiconductors remain the most consequential front. For years, the US tried to deny China access to advanced chipmaking equipment and top-end processors. Yet Chinese firms have spent heavily on domestic substitutes. SMIC is now one of the world’s largest chipmakers, while Huawei, Alibaba and Baidu are pursuing their own processors to reduce dependence on Nvidia. Analysts cited in the source material say China’s most advanced chips may still trail Western rivals, but the gap is narrowing.

The bigger challenge for the West may come from manufacturing equipment itself. China has long depended on ASML, the Dutch company that dominates advanced lithography systems used to make cutting-edge chips. After US pressure limited exports to China, reports emerged at the end of July that a state-backed Chinese company is developing its own lithography tools. The aim is to deliver the first five systems this year and around 20 by 2027. Even if they remain below ASML’s standard, the effort marks a meaningful step towards self-sufficiency and helped drive ASML’s share price lower.

Beijing’s 15th five-year plan, adopted in March, makes the logic explicit. Artificial intelligence, chips, robotics, quantum computing and space technology sit at the centre of China’s industrial strategy through 2030. By contrast, renewable energy and electric mobility, which were more prominent in the previous plan, now receive less emphasis because China believes those sectors are already well established.

The contest is also turning into a fight for people, not just hardware. Taiwan has opened investigations into alleged illegal recruitment of engineers by Chinese-linked companies, with authorities saying dozens of locations were searched and more than 100 people questioned. The island’s officials see the effort as a direct attempt to erode its advantage in advanced chip design and high-end manufacturing. In that sense, the competition is no longer just about who builds the best model or the fastest chip. It is about who controls the talent, the know-how and the industrial base that make technological leadership possible.