news

like basic resources such as water, electricity, and gas, computing power also requires a unified calculation standard. on september 23, jiuzhang yunji gave its own solution: computing power package, that is, using "degree" as the computing power calculation unit, providing an integrated computing power resource package. 1 degree of computing power includes 312tflops*1 hour of gpu (graphics processing unit) computing power, network, storage, security services, intelligent computing operating system, and operation management and other services. from traditional time-based billing to volume-based billing, it seems to be a change at the sales level, but it is not only related to sales. for computing power users, it can be purchased flexibly and the threshold is lower. for jiuzhang yunji, it involves technical support for the intelligent computing operating system. will this move boost revenue? as of press time, people related to jiuzhang yunji have not yet responded.

"in the next few decades, we will be able to do things that seemed like magic to our grandparents. our abilities have increased not because of genetic changes, but because we benefit from a society whose infrastructure is smarter and more capable than any of us," said openai ceo sam altman in a blog post published in the early morning of september 23. when alibaba cloud announced another price cut not long ago, alibaba cloud cto zhou jingren also repeatedly emphasized the role of ai infrastructure in the flourishing of ai applications.

jiuzhang cloud also focused on ai infrastructure. on september 23, it launched a computing power package product and officially launched the alaya new cloud intelligent computing dedicated cloud service. alaya new is an intelligent operating system launched by jiuzhang cloud in april. the computing power package manages and divides the gpu computing power resource pool through this system, and breaks down the computing power into small computing power units. the computing power package uses "degrees" as the basic unit of measurement, charges according to the amount of computing, and adopts a pre-charge and real-time deduction model.

"if we imagine that gpus, hardware, and related capabilities are airplanes, the travel needs of a family, individual, and organization are diverse, and the role of an airplane may change from passenger transport to cargo transport. this requires a middle layer to meet supply and demand, and the middle layer is the airline." fang lei, chairman of jiuzhang yunji, gave an example, "airlines provide economy class, business class and other services that adapt to the market through efficient operations. airlines have relatively fixed supply, but are connected to changing demands, playing a connecting role in the middle." obviously, what jiuzhang yunji wants to do is to be a computing power "airline."

shang mingdong, co-founder and cto of jiuzhang yunji, talked about the cost advantage of computing power packages and said, "if you migrate to a computing power package, the price is guaranteed to be no higher than bare metal. based on the comparison of monthly costs and task volumes, compared with bare metal, the unit price of computing power packages is cheaper."

the "china intelligent computing center operating system market research white paper 2024" brought by zhu xiaowen, consulting director of sullivan, shows that the scale of china's core artificial intelligence industry is expected to reach 731.6 billion yuan in 2025, which will drive the scale of related industries to about 5 trillion yuan. during the "14th five-year plan" period, if the intelligent computing center reaches an application level of 80%, its investment is expected to increase the core artificial intelligence industry by 2.9-3.4 times and related industries by 36-42 times.

in the challenge section, zhu xiaowen mentioned standardization. "the slow progress of computing power standardization has led to a lack of unified computing power supply standards in the market. this situation makes it difficult for intelligent centers to implement standardized computing power pricing and supply models when providing services. different intelligent computing centers may adopt their own pricing methods and standards, which has caused trouble and inconvenience to downstream customers when purchasing and using computing power." however, people related to jiuzhang yunji did not disclose the impact of setting standards on commercialization.

beijing business daily reporter wei wei