Supervisory board chair says company will focus on connectivity and computing, building advantage through clusters and supernodes

Huawei's AI computing system Atlas 950 SuperPod is on display at the venue for the 2026 World Artificial Intelligence Conference in Shanghai, China, on Wednesday (local time). [Reuters]
Huawei's AI computing system Atlas 950 SuperPod is on display at the venue for the 2026 World Artificial Intelligence Conference in Shanghai, China, on Wednesday (local time). [Reuters]

Huawei, the Chinese telecommunications equipment maker at the forefront of the US-China technology rivalry, has set its sights on becoming the Nvidia of the information and communications technology and computing sectors. The company's strategy is to overcome its disadvantage in advanced chip manufacturing — a result of US semiconductor export restrictions — by linking thousands of AI chips into a single computing pool through "supernode" and cluster technologies.

Guo Ping, chair of Huawei's supervisory board — the company's highest oversight body — made the remarks in a recent conversation with new employees, according to Chinese media outlets including Yicai and Sina Tech on Wednesday (local time). He was responding to a question about whether Huawei should develop its own large language model as part of its AI industry strategy.

"In the ICT business and computing space, Huawei's goal is to become Nvidia," Guo said. "Huawei's core advantage lies not in having its own in-house LLM, but in supporting customers to build superior LLMs."

The vision is to ensure that diverse LLMs from around the world can run efficiently on Huawei's computing infrastructure — including its Ascend AI processors, Kunpeng server processors, supernodes and clusters. Guo said Huawei is concentrating its capabilities on connectivity and computing and has no plans to expand into new business areas.

When asked how Huawei's competitive advantages compare with Nvidia's — whose moat lies in advanced manufacturing processes, hardware-software integration and the CUDA software platform ecosystem — Guo pointed to supernodes and similar technologies.

A supernode uses high-speed interconnect technology to link as many as thousands of AI chips into a single computing pool. While China lags behind the United States at the level of individual chips, the idea is that it can compensate through system-level advantages in linking those chips together.

"Because Huawei faces constraints in advanced manufacturing processes, we have an even greater need to integrate design and manufacturing," Guo said. "We must leverage our overall strengths to make up for technological shortcomings." He also referenced Huawei's "Tao's Law," which the company has proposed as a framework for overcoming the limits of Moore's Law.

Guo then cited remarks by DeepSeek founder Liang Wenfeng, who said AI advances are rapidly eroding the moat of Nvidia's CUDA ecosystem. "Huawei will secure an overall advantage through the development of clusters and supernodes," he said.

On the broader AI outlook, Guo said AI could prove to be the final technological revolution in human history, and that companies must secure a competitive edge within it.

He also said that while a small number of top-tier talents will build AI in the future, the vast majority of people will use it — and that the ability to ask the right questions will be a key skill in an AI-driven world, drawing growing attention to graduates in literature and philosophy.

Guo assessed that China trails in computing power — one of the three pillars of AI — but holds an advantage in data and is roughly on par with the United States in talent.


mokiya@heraldcorp.com