xi's moments
Home | Technology

Country striving to build up computing muscle

By FAN FEIFEI | China Daily | Updated: 2023-02-28 07:34

Technicians check equipment at a computing center in Nanchang, Jiangxi province, in September. WAN XIANG/XINHUA

China's computing power sector is expected to witness speedy growth as tech companies are scrambling to roll out ChatGPT-like artificial intelligence chatbots, which necessitates higher requirements for computing capacity in the process of large language model training and operation, experts said.

Computing power, or the ability to process data, is widely regarded as a vital foundation for bolstering the development of the digital economy and a new engine to unleash the potential of data as a factor of production, they added.

According to investment bank China International Capital Corp, the computing power consumed by ChatGPT training is about 3,640 PF-days — or petaflop days — which means 1 quadrillion calculations per second are needed and will run continuously for 3,640 days.

The market scale of China's computing power industry exceeded 1.5 trillion yuan ($215.4 billion) in 2021, with revenue from cloud computing surpassing 300 billion yuan and internet data center services over 150 billion yuan, said the China Academy of Information and Communications Technology, a government think tank.

Chinese enterprises have ramped up efforts to build intelligent computing centers and bolster the development of ChatGPT-like products and services by offering computing power support.

Supercomputer manufacturer Dawning Information Industry Co, also known as Sugon, participated in China's first integrated computing power trading platform, which was unveiled in Yinchuan, Ningxia Hui autonomous region, on Friday.

Cao Zhennan, vice-president of Sugon, said the company has accumulated advantages in large-scale computing infrastructure and computing power resource scheduling. It will optimize computing structure on the trading platform, and improve computing supply capacities for artificial intelligence and scientific engineering projects.

The platform involves a batch of companies and institutions such as Alibaba Cloud, Huawei Technologies and SenseTime. It is part of the country's broader push to channel more computing resources from eastern regions to less-developed but resource-rich western ones.

Computing power is essential as generative AI is currently making disruptive breakthroughs, said Daniel Zhang, chairman and CEO of Alibaba Group, in an earnings call with investors on Thursday night, adding that the company will continue to build its own large-scale pre-training models.

"We are also ready to capture market opportunities to provide computing power for various generative AI models and their applications," Zhang said.

Alibaba Cloud, the cloud computing arm of Alibaba, has launched the world's largest intelligent computing center in Zhangjiakou, Hebei province.

The center will provide intelligent computing services for AI-powered large-scale model training, autonomous driving, spatial geography and other AI-related frontier applications, the company said.

Zou Debao, deputy general manager of the AI industry research institute from market research firm CCID Consulting, said AI model training has huge demand for computing power, and with the increase of training intensity and computing complexity, requirements for computational accuracy are also gradually rising.

The establishment of intelligent computing centers will improve efficiency in the use of computing resources and accelerate enterprises' digitalization push, said Xiang Ligang, director-general of the Information Consumption Alliance, a telecom industry association.

Global Edition
BACK TO THE TOP
Copyright 1995 - . All rights reserved. The content (including but not limited to text, photo, multimedia information, etc) published in this site belongs to China Daily Information Co (CDIC). Without written authorization from CDIC, such content shall not be republished or used in any form. Note: Browsers with 1024*768 or higher resolution are suggested for this site.
License for publishing multimedia online 0108263

Registration Number: 130349