Nvidia absent from China's biggest AI event, Huawei takes center stage with new Ascend AI chips

Nvidia's absence from China's largest AI event in Shanghai has sparked concerns about the company's presence in the country. Chinese rivals Huawei, Moore Threads, and Yunsilicon showcased their AI chips, while Nvidia's competitors such as Tencent and Alibaba demonstrated AI applications in robotics, smart glasses, and translation apps. Huawei unveiled its Atlas 900 A3 SuperPoD system, which links 384 Ascend AI chips together. The US chipmaker's absence comes amid renewed hopes of selling its H20 chips to China again.

Title: Nvidia's Absence from Shanghai AI Event Sparks Concerns Over China Presence

Nvidia's absence from China's largest AI event in Shanghai has raised eyebrows, particularly as Chinese rivals like Huawei, Moore Threads, and Yunsilicon showcased their AI chips. The event saw competitors such as Tencent and Alibaba demonstrating AI applications in robotics, smart glasses, and translation apps. Notably, Huawei unveiled its Atlas 900 A3 SuperPoD system, which links 384 Ascend AI chips together. This event coincides with renewed hopes of Nvidia resuming sales of its H20 chips in China, following reports of potential new licenses from the U.S. government [4].

Nvidia's absence from the event has sparked concerns about the company's presence in the country. The U.S. chipmaker has been facing export control issues, with at least $1 billion of advanced AI chips reportedly smuggled into China in recent months [1]. Despite these challenges, Nvidia's CEO Jensen Huang announced last week that the company expects to receive assurances from U.S. authorities to resume sales of its H20 chip in China [4].

Huawei, on the other hand, has been making significant strides in the AI segment. The company showcased its CloudMatrix rack-scale cluster for the first time, featuring 384 Ascend 910C chips connected in an "all-to-all topology" configuration. This AI cluster reportedly delivers 300 PetaFLOPS of BF16 computing, nearly double the performance of NVIDIA's GB200 NVL72. However, the CloudMatrix 384 consumes significantly more power, with estimates suggesting it uses nearly four times as much power as the GB200 NVL72 [2].

The event underscores the growing competition in the AI chip market, with Chinese firms increasingly developing their own technologies. While Nvidia's absence is notable, it also highlights the potential for other players to fill the void. As the AI market continues to grow, the ability to produce and sell AI chips will be a key factor in determining market share.

References:
1. [https://www.tahawultech.com/home-slide/nvidia-ai-chips-reportedly-smuggled-into-china/](https://www.tahawultech.com/home-slide/nvidia-ai-chips-reportedly-smuggled-into-china/)
2. [https://wccftech.com/huawei-showcases-its-nvidia-gb200-equivalent-cloudmatrix-ai-cluster-for-the-first-time-to-the-public/](https://wccftech.com/huawei-showcases-its-nvidia-gb200-equivalent-cloudmatrix-ai-cluster-for-the-first-time-to-the-public/)
3. [https://notebookspec.com/web/831796-huawei-cloudmatrix-ai-cluster](https://notebookspec.com/web/831796-huawei-cloudmatrix-ai-cluster)
4. [https://www.rcrwireless.com/20250721/ai-infrastructure/nvidia-h20-chip](https://www.rcrwireless.com/20250721/ai-infrastructure/nvidia-h20-chip)