China Forms AI Alliances to Counter U.S. Chip Restrictions at Shanghai Conference

China’s leading artificial intelligence firms have unveiled two major alliances to build a domestic AI ecosystem and lessen reliance on foreign technology amid U.S. export controls on advanced Nvidia (NASDAQ: NVDA) chips. The announcements came during the World Artificial Intelligence Conference in Shanghai, which concluded Monday.

The Model-Chip Ecosystem Innovation Alliance connects large language model (LLM) developers with Chinese GPU manufacturers including Huawei, Biren, Moore Threads, and Enflame. These companies, affected by U.S. sanctions, aim to strengthen cooperation across chips, models, and infrastructure. StepFun, an LLM developer, led the initiative.

A second group, the Shanghai General Chamber of Commerce AI Committee, seeks to accelerate AI integration across industries. Members include SenseTime—pivoting from facial recognition to LLMs—alongside MiniMax, Metax, and Iluvatar CoreX.

The conference showcased advanced AI hardware, notably Huawei’s CloudMatrix 384, powered by 384 910C chips. U.S. research firm SemiAnalysis reported it surpasses Nvidia’s GB200 NVL72 in certain benchmarks through innovative system-level design and clustering techniques. Metax also revealed an AI supernode using 128 C550 chips for large-scale data centers.

Consumer-focused AI products were highlighted as well. Tencent introduced the open-source Hunyuan3D World Model 1.0, enabling 3D environment generation from text or images. Baidu (NASDAQ: BIDU) unveiled next-gen “digital human” technology capable of cloning voice and body language from minimal footage, aimed at virtual livestreaming. Alibaba (NYSE: BABA) presented Quark AI Glasses, powered by its Qwen model, offering navigation and QR-enabled Alipay payments, slated for release in late 2025.

These developments underscore China’s push to innovate domestically and compete globally in AI amid intensifying tech restrictions.