[News] Huawei Led Chinese AI Army to Counter U.S. Curbs with New Model-Chip Alliance

Just days after Trump unveiled his AI Action Plan on July 23—aimed at expanding U.S. AI globally to curb dependence on Chinese chips and models—China fired back by launching two industry alliances to strengthen its own AI ecosystem. The push, led by Huawei and other domestic tech giants, was highlighted by Reuters and CNBC.

The announcements—launching the Model-Chip Ecosystem Innovation Alliance and the Shanghai AI Committee—were made during the three-day World Artificial Intelligence Conference (WAIC) in Shanghai, which concluded on Monday, according to the reports.

Model-Chip Ecosystem Innovation Alliance

The reports suggests that the Model-Chip Ecosystem Innovation Alliance brings together China’s leading LLM developers and AI chipmakers to create a unified tech stack. Members include domestic GPU heavyweight Huawei and rising players like Biren and Moore Threads—all facing U.S. sanctions that restrict access to advanced technologies, the reports add.

It is worth noting that Huawei’s CloudMatrix 384 system was shown for the first time at the 2025 WAIC. According to chinastarmarket.cn, Huawei’s 384 stole the spotlight, as it is the first of its kind to connect 384 Ascend NPUs and 192 Kunpeng CPUs through the company’s high-speed MatrixLink network, forming a massive AI super server with 300 PFLOPs of compute power.

As highlighted by an earlier Tom’s Hardware report, while a single Ascend 910C chip delivers just one-third the performance of NVIDIA’s Blackwell, Huawei makes up for it by packing far more chips into each system. As the report highlights, this strategy pushes the CloudMatrix 384 to around 300 PFLOPs of dense BF16 compute—nearly double the 180 PFLOPs offered by NVIDIA’s GB200 NVL72.

Shanghai General Chamber of Commerce AI Committee

Meanwhile, the second group announced at WAIC, the AI Committee of the Shanghai General Chamber of Commerce, aims to accelerate AI’s integration into industry. Key members include SenseTime, LLM developer StepFun, MiniMax, and chipmakers MetaX and Iluvatar CoreX, the Reuters report suggests.

Read more

• [News] Huawei to Unveil CloudMatrix 384, Rumored to Deliver 2× NVIDIA GB200 NVL72 Throughput

(Photo credit: Huawei)