Chinese AI Firms Forge Alliances to Build Domestic Ecosystem Amid U.S. Sanctions

In a strategic move to mitigate the impact of U.S. export restrictions on advanced technology, particularly Nvidia chipsets, China's artificial intelligence (AI) companies have announced the formation of two significant industry alliances aimed at fostering a self-sufficient domestic ecosystem. The announcements coincided with the three-day World Artificial Intelligence Conference held in Shanghai, which concluded on July 28, 2025.

The first alliance, dubbed the "Model-Chip Ecosystem Innovation Alliance," unites Chinese developers of large language models (LLMs) with AI chip manufacturers. Zhao Lidong, CEO of Enflame, one of the participating chipmakers, emphasized the importance of this initiative, stating, "This is an innovative ecosystem that connects the complete technology chain from chips to models to infrastructure." Other prominent GPU manufacturers in this coalition include Huawei, Biren, and Moore Threads, all of which have faced U.S. sanctions that hinder their ability to procure advanced technologies reliant on U.S. intellectual property.

The second alliance, facilitated by the Shanghai General Chamber of Commerce AI Committee, aims to enhance the integration of AI technology within various industries. Key participants include SenseTime, which has shifted its focus from facial recognition to LLM development, as well as other LLM developers such as StepFun and MiniMax, alongside chipmakers Metax and Iluvatar CoreX.

At the World Artificial Intelligence Conference, several innovative products were showcased, including Huawei's CloudMatrix 384, which reportedly outperforms Nvidia's GB200 NVL72 in certain metrics, according to SemiAnalysis, a U.S.-based research firm. Huawei's design capabilities have allowed it to leverage a higher number of chips and innovative system-level strategies to counterbalance individual chip performance limitations. Additionally, at least six other Chinese computing firms presented similar clustering chip technologies, with Metax unveiling an AI supernode featuring 128 C550 chips suited for large-scale, liquid-cooled data centers.

Other notable innovations revealed at the conference included Tencent's Hunyuan3D World Model 1.0, which enables users to generate interactive 3D environments using text or image prompts, and Baidu's next-generation "digital human" technology aimed at aiding businesses in creating virtual livestreamers. This technology boasts cloning capabilities, replicating a human's voice, tone, and body language from merely ten minutes of sample footage. Furthermore, Alibaba announced its upcoming Quark AI Glasses, powered by its Qwen AI model, set for release in China by the end of 2025. These glasses will allow users to navigate using the company's mapping services and facilitate payments through Alipay via voice commands.

As the landscape of AI technology evolves in China, the establishment of these alliances signifies a crucial step toward achieving technological independence, particularly in light of ongoing geopolitical tensions and trade barriers. Experts warn, however, that while these initiatives may bolster domestic capabilities, the reliance on home-grown technologies must be accompanied by rigorous research and development to ensure competitiveness on the global stage.

In conclusion, the formation of these alliances not only reflects China's response to external pressures but also underscores the broader implications for the global AI industry. As nations increasingly prioritize self-sufficiency in critical technologies, the competition among AI firms is set to intensify, potentially reshaping the global technological landscape in the years to come.