The China AI strategy is making its mark. Artificial Intelligence research in the country is rapidly evolving to reduce reliance on foreign technologies and boost domestic innovation. As geopolitical tensions over technology persist, China’s artificial intelligence sector is taking decisive steps to reduce its reliance on foreign hardware, particularly advanced chips from US-based companies such as Nvidia.
With growing global competition, China's AI strategy now focuses on smart infrastructure and edge computing. At the recent World Artificial Intelligence Conference (WAIC) in Shanghai, Chinese tech firms unveiled two major alliances aimed at strengthening the local AI ecosystem and addressing challenges caused by ongoing US export restrictions.
The Model-Chip Ecosystem Innovation Alliance, one of the key initiatives announced, connects local large language model (LLM) developers and local semiconductor manufacturers.
The alliance includes many of the most prominent players in the domestic space, including Huawei, Biren, Enflame, and Moore Threads, all of which were impacted by US sanctions, which cut them off from high-end GPUs.
Zhao Lidong, CEO of Enflame, described the alliance as a comprehensive ecosystem, an aspect of which connects the development of chips to AI models and infrastructure, thereby setting the stage for innovation under constrained conditions.
A second consortium, the Shanghai General Chamber of Commerce AI Committee, focuses on applying AI across industrial sectors. Members include SenseTime, which has pivoted from facial recognition to LLMs after facing sanctions, alongside LLM firms StepFun and MiniMax, and chipmakers Metax and Iluvatar CoreX.
Huawei made headlines at WAIC by unveiling its CloudMatrix 384, a high-performance AI system powered by 384 of its latest 910C chips. Analysts at SemiAnalysis report that the system can outperform Nvidia’s leading GB200 NVL72 in specific benchmarks, mainly due to Huawei’s system integration and innovative clustering approach.
Huawei CloudMatrix is playing a key role in supporting China’s AI ambitions through scalable cloud services. Other Chinese firms also demonstrated similar cluster-based systems. Metax showcased an AI supernode powered by 128 C550 chips, designed for large-scale, liquid-cooled data centers.
Tightening US Chip Sanctions have prompted Chinese tech firms to seek local alternatives and intensify R&D efforts. These advancements reflect China’s broader move toward scaling domestic data-processing infrastructure to compensate for hardware shortages caused by export restrictions.
Also Read: China’s Premier Li Qiang Calls for Global AI Cooperation at WAIC 2025
Alongside enterprise hardware, Chinese tech giants also revealed several AI-driven consumer products. Alibaba introduced its Quark AI Glasses, which utilize the Qwen model to provide voice-enabled navigation and QR code scanning capabilities. The glasses are set for release by late 2025.
Tencent unveiled Hunyuan3D World Model 1.0, an open-source platform enabling users to generate 3D environments using text or images. Baidu, meanwhile, showcased its advanced 'digital human' technology, which can replicate a real person’s voice and gestures using minimal training data.
At the Meizu AI Conference, industry leaders showcased new AI products built on homegrown platforms. Chinese Semiconductor Alliances are forming to develop a self-sustaining chip ecosystem amid international restrictions.
These developments signal China’s commitment to technological self-sufficiency as the country continues to face pressure from the US regarding access to AI and semiconductors.
By forming industry alliances and accelerating domestic innovation, Chinese firms are laying the groundwork for a resilient, homegrown AI landscape that is designed to thrive even in the face of external constraints.