Computing power leasing broke out across the board! The key track of AI computing power, the core la

Mondo Technology Updated on 2024-02-25

Computing power, that is, the amount of information and data that the device can process per second according to the change of the internal state, the development process of its carrier has gone through the era of abacus and mechanical calculators to the server based on the Internet basic platform. In the past two decades, the richness of computing power carriers has been unprecedentedly improved, and it has shown a trend of diversification.

The supply of computing power can be divided into three main types: general computing power, intelligent computing power and supercomputing computing power. The core of the realization of this computing power lies in various computing chips, such as CPUs, GPUs, FPGAs, and ASICs, which are embedded in computers, servers, high-performance computing clusters, and various intelligent terminals. Whether it is the processing of massive data or various digital applications, the processing and calculation of these computing power are inseparable. Artificial intelligence

With the rapid development of AI large models in 2023, the demand for high-end intelligent computing power has soared exponentially. The shortage of computing resources is driving the supply of computing power towards the era of sharing. In this context, the computing power leasing industry has ushered in an unprecedented boom period.

In the face of the surge in computing power demand brought about by the wide application of artificial intelligence technology, especially large models, many enterprises are feeling great pressure.

Intelligent computing power has become the core force driving AI large model training, and high-performance GPUs are the key to meeting this demand, and the imbalance between supply and demand has led to a rise in the highest quality and a longer delivery cycle. This situation has caused many enterprises to encounter bottlenecks on the road to the pursuit of computing power.

The emergence of the computing power leasing model provides a new solution for enterprises. Thanks to the increasing maturity and popularity of cloud computing technology, computing power leasing has become an ideal choice for providing users with powerful computing power in a flexible and on-demand manner. It allows users to access the required computing power through the cloud anytime and anywhere, which not only greatly enhances the scalability and flexibility of computing power, but also realizes true pay-as-you-go, saving a lot of costs for enterprises.

For those companies with computing power needs, renting cloud computing power can not only significantly reduce hardware costs, but also improve cost control. In addition, the migration of computing power to the cloud also lays the foundation for efficient computing power scheduling.

For enterprises focusing on the research and development of small models in the vertical industry, the computing power leasing model is a powerful tool to lower the threshold for the research and development of large models. They no longer need to invest heavily in buying and maintaining expensive servers, and they can easily access the computing center through a browser and enjoy powerful computing services.

With its flexibility, scalability, and cost-effectiveness, the computing power leasing model is expected to become the best choice for computing power supply in the future.

Fact**: The commercial essence of qubit AI computing power leasing is actually to provide a service for software developers who need large model training to lease GPU computing power to manufacturers with abundant GPU resources.

In this business model, the lessor's main profit** is the difference between rental income and operating costs.

Among them, operating costs cover many aspects, including depreciation expenses of fixed assets, daily operating expenses of data centers, rental of IDC equipment rooms, and personnel costs.

Taking 2023 as an example, if the rental equipment is used as the H800 server, the annual rent** of the computing power leasing industry will remain at the level of about 120,000 yuan per year. Based on such a system, the cycle of computing power leasing is about 17 months.

With the continuous increase in the cost of computing power resource allocation, the computing power leasing business has entered the track of rapid development. By effectively integrating and encapsulating computing power, storage, and network resources, and providing them to customers in the form of services (e.g., through API interfaces), the enterprise can significantly reduce the initial investment and subsequent operating costs.

The upstream of the computing power leasing industry chain is the leading computing power provider, the midstream is the cloud vendor, and the downstream is the AI application organization and enterprise.

Among them, NVIDIA is the world's largest supplier of high-performance computing power graphics cards, almost monopolizing the entire AI server chip market.

Midstream cloud vendors include Alibaba, Tencent, China Mobile, China Unicom, and China Telecom. Cloud vendors need to provide downstream users with computing power leasing business-related resources, including not only cabinets, bandwidth, and power, but more importantly, lease appropriate computing power (hardware configuration + software environment) according to user needs.

Diagram of the computing power leasing industry chain:

In addition, whether the traditional business can form a synergistic effect with the computing power leasing business, and it should be noted that the computing power leasing industry is affected by geopolitics, policies, laws, etc.

In addition, the manufacturers that laid out in the computing service industry in the early stage have resource advantages and first-mover advantages in the field of computing power leasing. The main domestic server manufacturers include Inspur Information, Sugon, Industrial Fortune Union, Tuowei Information, etc.; Enterprises related to the layout of computing power leasing services include Hongbo Co., Ltd., Zhongbei Communication, Lotus Health, Runjian Co., Ltd., iSoftStone, Qingyun Technology, Winner Technology, etc.

At present, due to the mismatch between the supply and demand of computing power, the computing power leasing business presents the characteristics of a seller's market. In this context, the highest level of computing power leasing is relatively high, and the lessor has greater bargaining power in the market. However, it also provides a huge opportunity for the development of the computing power leasing industry.

With the continuous advancement of technology, especially the rapid development of artificial intelligence, big data and other fields, the demand for computing power will continue to grow. Whether it is autonomous driving, robotics, smart medical care, entertainment creation and other fields, they all need strong computing power support. As a flexible and efficient way to obtain computing power, computing power leasing will become more and more favored in these fields.

At the same time, with the continuous growth of market demand, the scale of the computing power leasing industry will also continue to expand. More computing power providers will enter the market and provide more computing resources; Cloud vendors will also continue to improve their computing power leasing services to meet the diverse needs of users. Downstream AI application organizations and enterprises will also rely more on computing power leasing to promote the rapid development of the industry.

In addition, with the continuous development of the computing power leasing industry, its business model will also continue to innovate and improve. For example, cloud vendors can achieve optimal allocation and efficient utilization of computing power through in-depth cooperation with computing power providers; At the same time, they can also provide customized computing power rental services to meet the specific needs of users. These innovations will promote the development of the computing power leasing industry to a higher level.

Pay attention to [Leqing Think Tank] and gain insight into the industrial pattern!

Related Pages