Chinese tech giant Huawei has launched a global challenge in the artificial intelligence (AI) battle by shifting its focus from nanometer chip manufacturing processes to computing power and efficiency. Zhang Pingan, CEO and Executive Director of Huawei Cloud, announced that the company believes its computing power is three times more efficient than rival NVIDIA’s AI chips.
In his statement, Zhang Pingan emphasized that chip manufacturing processes like 5nm and 7nm, which are often the focus of the tech world, are not actually the key to AI. According to the CEO, customers’ primary demand is for high-quality computing results, and the key drivers for this are computing power and efficiency.
Huawei stands out with its speed and reliability
Huawei’s strategy represents not only a necessity but also a philosophical shift for the company, which has struggled to access the most advanced chip manufacturing technologies (like 5nm), particularly due to US export restrictions. Zhang Pingan summarized this situation with these striking words:
Chip processes (like 5nm and 7nm) are not the key. What customers really need is high-quality computing results. Huawei Cloud Service has revolutionized computing power and efficiency, with throughput reaching three times that of Nvidia’s H20 chip.
This statement means the company claims to offer better throughput than its competitor’s current-generation products despite hardware limitations thanks to software optimization and system architecture (e.g., the Ascend ecosystem).
The Huawei executive supported this efficiency increase with a concrete example: Huawei Cloud Service is capable of generating 2,400 tokens per second on a single card with a latency of just 50 milliseconds.
Zhang also noted that the company has increased the reliability and coverage of its cloud platform, Ascend Cloud Service, over the years. The service confirmed that it works effectively with Huawei’s own large language model, Pangu, as well as third-party Large Language Models (LLMs) such as DeepSeek and Kimi. Zhang stated that these collaborations are acceptable, adding that soon all LLMs will run faster and better on Ascend Cloud technologies.
Huawei’s breakthrough aligns with China’s ambition to become a global hub for AI and computing despite international sanctions. The company notes that Huawei Cloud’s 756 consecutive days of zero major incidents have made a valuable contribution to this segment.
