Huawei Says Chinese AI Must Accelerate Development to Experience Frontier Risks US Labs Already Face

Huawei Calls for Accelerated Chinese AI Development Amid US Safety Debates

Huawei's rotating chairman Eric Xu made a striking argument at the company's annual Connect conference in Shanghai, stating that Chinese AI developers may not yet be advanced enough to experience the safety risks reported by leading US firms1

1

. Speaking to reporters, Xu suggested that Chinese AI labs should continue developing more powerful models while balancing innovation against AI risks, directly contrasting recent calls from OpenAI and Anthropic for coordinated slowdowns in frontier AI development2

2

.

"If we look at the mainstream AI model providers in the United States, because they have massive computing power, maybe only they themselves know where they are in terms of the level of their models," Xu explained1

1

. "Maybe the type of risk from AI that they can feel, they can perceive, is something that people in China or AI model providers in China cannot."

Diverging Approaches to AI Safety Concerns

The timing of Xu's remarks highlights a fundamental split in how the world's two largest economies approach AI safety concerns. While alarm over AI safety has intensified in the United States, where leading developers and researchers have warned that increasingly autonomous systems could bypass safeguards or become difficult to control, China has taken a different approach4

4

. Beijing generally treats advanced AI as a powerful but manageable technology and is pressing ahead with rapid deployment while developing mandatory standards, security assessments and other safeguards against risks including autonomous AI agents circumventing controls or attacking outside systems1

1

.

OpenAI this week said it would regularly disclose unexpected or unauthorized model behavior, while Anthropic CEO Dario Amodei has called for slowing development to allow safeguards to catch up1

1

. Some Chinese commentators have portrayed proposals to "pace" frontier AI development as an attempt to constrain rivals that remain behind leading US labs, rather than simply as a response to AI safety concerns3

3

.

Huawei's Push for Self-Sufficiency Amid US Export Controls

Huawei is China's main domestic supplier of the computing infrastructure needed to train advanced AI models. Its outsize role in China's $50 billion AI chip market stems directly from US export controls since 2023 that restricted Chinese companies' access to cutting-edge chips from US giant Nvidia1

1

. Xu acknowledged Huawei remained "not as advanced" as leading US rivals but said the unpredictable export controls had pushed Chinese customers towards domestic technology2

2

.

"We can't accept a destiny that we cannot control," Xu stated2

2

. "No matter if it's the Chinese government or Huawei, we are pushing for full self-sufficiency for chips and the entire supply chain around semiconductors." The company announced that its Ascend 960 DT chip, which targets AI model training, would launch in the first quarter of 2027, while its inference chip would come in the third quarter, delivering twice the computing performance of its current flagship product2

2

.

Ambitious Projections for Autonomous AI Agents

Huawei itself expects autonomous AI agents to proliferate rapidly across the global AI landscape. The company forecast this week that agents—AI systems that can plan tasks, make decisions and use software or other tools with limited human supervision—would account for more than 90% of global AI processing traffic by 20351

1

. Huawei has projected as many as 900 billion active agents by then and identified agent security and privacy as crucial technologies4

4

.

Xu said at least six or seven Chinese AI labs were aiming over the next two years to train models containing 10 trillion to 40 trillion parameters, a significant increase from current model sizes in China2

2

. However, Xu said separately that Huawei could not produce enough AI computing equipment to satisfy domestic AI computing demand1

1

.

Technical Advances and Competitive Positioning

At the heart of Huawei's effort to challenge Nvidia is its networking technology, which it says allows it to build larger clusters than its US rival without compromising bandwidth and stability2

2

. Huawei's data centre product built around the Ascend chips will connect as many as 4,096 processors, compared with 1,024 in the previous generation, allowing customers to train and run larger AI models2

2

. This approach is central to Huawei's effort to compensate for the weaker performance of its individual AI hardware, as US export controls have prevented it from using leading-edge contractors such as Taiwan's TSMC2

2

.

Heng Liao, chief scientist for Huawei's semiconductor division, said changes in how increasingly large and complex AI models are trained were also eroding the advantage of Nvidia's CUDA platform, which has long helped lock developers into the US chipmaker's graphics processing units2

2

. "CUDA is not as important as it was two years ago," Liao said, pointing to the emergence of new programming tools and approaches for AI model training2

2

. Xu estimated that Ascend chips had overtaken Nvidia in market share in China while acknowledging that reliable estimates were difficult to obtain2

2

.