According to Fast Technology on October 4, in the view of Zhang Pingan, Huawei's executive director and CEO of Huawei Cloud, chip process is not the core, what customers really need is high-quality computing results.
Recently, Zhang Pingan publicly stated that Huawei Cloud Service has achieved a breakthrough in computing power and efficiency, and its production efficiency has reached three times that of Nvidia's H20 chip.
"The chip process (such as 5nm and 7nm) is not the core. What customers really need is high-quality computing results." Zhang Pingan introduced that through technological innovation, Huawei Cloud Service has achieved the ability to generate 2,400 tokens per second on a single card with a latency of 50 milliseconds.
Currently, Huawei's Ascend Cloud service not only supports our own large-scale Pangu models, but also fully supports third-party models such as DeepSeek and Kimi. "We hope that all large-scale models will run faster and better on Ascend Cloud," said Zhang Pingan.
Zhang Pingan pointed out that China's "computing power hub" is gradually becoming an AI computing power center for global customers. During Huawei's overseas expansion, it discovered that while intelligent computing centers in China all utilize liquid cooling technology, liquid cooling data centers overseas are still relatively rare. Renovating overseas data centers is not only time-consuming, but also lacks sufficient fiber optic network bandwidth.
Zhang Pingan finally emphasized that according to regulations, every cloud vendor must publicly disclose any major incidents online, and Huawei Cloud has maintained a "zero major incident record" for 756 consecutive days. "We are confident that we can maintain this achievement," he said.
Recently, Zhang Pingan publicly stated that Huawei Cloud Service has achieved a breakthrough in computing power and efficiency, and its production efficiency has reached three times that of Nvidia's H20 chip.
"The chip process (such as 5nm and 7nm) is not the core. What customers really need is high-quality computing results." Zhang Pingan introduced that through technological innovation, Huawei Cloud Service has achieved the ability to generate 2,400 tokens per second on a single card with a latency of 50 milliseconds.
Currently, Huawei's Ascend Cloud service not only supports our own large-scale Pangu models, but also fully supports third-party models such as DeepSeek and Kimi. "We hope that all large-scale models will run faster and better on Ascend Cloud," said Zhang Pingan.
Zhang Pingan pointed out that China's "computing power hub" is gradually becoming an AI computing power center for global customers. During Huawei's overseas expansion, it discovered that while intelligent computing centers in China all utilize liquid cooling technology, liquid cooling data centers overseas are still relatively rare. Renovating overseas data centers is not only time-consuming, but also lacks sufficient fiber optic network bandwidth.
Zhang Pingan finally emphasized that according to regulations, every cloud vendor must publicly disclose any major incidents online, and Huawei Cloud has maintained a "zero major incident record" for 756 consecutive days. "We are confident that we can maintain this achievement," he said.
张平安：5、7nm并非核心 华为算力能力已超英伟达芯片3倍
快科技10月4日消息，在华为常务董事、华为云CEO张平安看来，芯片制程并非核心，客户真正需要的是优质的计算结果。近日，张平安公开表示，华为云服务在算力效能方面实现突破，其生产效率已达到
news.mydrivers.com