"For Qingyun, the AI era will be a new opportunity, which will bring greater market demand and opportunities for 10 years than in the past 10 years. Lin Yuan, president of Qingyun Technology, said. In this round of technological progress of AI large models, the computing power required to support artificial intelligence has naturally become the focus of attention. Based on the insight into the demand for AI computing power, Tsing Cloud has released products and services such as Tsing Cloud AI Computing Service Platform and Tsing Cloud AI Computing Cloud.
With the development of large AI models, the importance of computing power has become increasingly prominent. Taking ChatGPT as an example, according to relevant survey data, the total computing power consumption of ChatGPT is about 3640pf-days, and 7 to 8 data centers with an investment scale of 3 billion and computing power of 500p are required to support its operation. It is worth noting that the computing power in AI training tasks is growing extremely fast, every 3It will double in 5 months.
While the demand for computing power is increasing, computing power has become the core force to promote the development of the digital economy. According to the latest data from the China Comprehensive Computing Power Index (2023), as of the end of June 2023, the total scale of computing power in China has reached 197eflops, and the total scale of data center racks in use in the country exceeds 7.6 million standard racks, ranking second in the world. The average annual growth rate of China's total computing power in the past five years is nearly 30%.
The development of the AI computing industry is both an opportunity and a challenge.
Lin Yuan believes that AI and large models are major productivity changes, and in this technology-driven market change, customer needs, computing power needs, application requirements, and underlying architecture are all changing. In the past decade, whether it is the mobile Internet for the C-end or the digitalization for the B-side, the application innovation is mainly based on the CPU. However, with the popularity of AI, almost all applications need to embrace AI technology or be modified by AI, which increases the demand for GPUs.
Not only that, under the trend of eastern data and western computing, AI applications in the future will pay more attention to cost-effectiveness while pursuing efficiency, so as to promote the development of computing power from the eastern coastal areas with low latency to the western regions with cheaper computing power, and the distribution of computing power will be more decentralized in the future. In addition, with the continuous development of AI and cloud-native technologies, customer services will become more complex, business formats will be more abundant, and challenges to technologies and platforms will also increase.
In particular, in the process of construction and operation of intelligent computing centers, how to consider the compliance of its design, how to achieve unified scheduling of multiple computing power such as supercomputing, intelligent computing, and general computing, how to take into account traditional applications, cloud-native applications, and AI applications at the same time, how to support heterogeneous CPUs and GPUs, how to provide efficient network and storage, and how to achieve efficient operation and maintenance and operation are all issues that need to be considered. It can be said that the operation of the intelligent computing center after the construction has become a key challenge.
Based on its own experience and practice, Qingyun Technology also gave the corresponding answer - AI computing power service platform, and provided a reference paradigm for the industry. Miao Hui, Product Manager of Qingyun Technology, said: "As a new model for the construction and operation of computing power centers, the AI computing service platform aims to manage AI infrastructure, provide multiple computing power scheduling, and provide intelligent computing power scheduling to ensure that computing power can be quickly established. ”
TsingCloud's AI computing power cloud is an AI computing power service that is mainly operated by Tsingcloud and in the form of ecological alliance. Compared with the traditional model, Qingyun AI computing cloud has certain advantages: from the customer's point of view, they need a complete solution, including machine room, computing power, scheduling platform, model driver, model, model service, and the above applicationsFrom the perspective of the best businessman, each layer requires professional technology and long-term accumulation and huge investment.
Lin Yuan believes that in the field of AI, giants will definitely appear, but these giants should not be an independent company, but a good ecological cooperation system, which requires all companies to work together to promote the progress and development of AI technology.