The surge in demand for computational power leasing has prompted numerous publicly listed companies to venture into this burgeoning sectorFollowing the initiation of 2024, a fascinating trend has emerged: not only have traditional software firms begun to dabble in computational leasing, but a variety of other businesses are also doing the same, marking a significant shift in the industrial landscape.

According to reports from industry research institutions, the potential revenue market for computational power leasing in China could soar to approximately 2.6 trillion yuan by 2026, with an annual growth rate exceeding 20%. This rapid expansion has attracted a diverse array of enterprises toward the computational leasing industryMany of these companies are new to the computational space, focusing predominantly on intelligent computing centers as they aim to establish footholds in this evolving field.

The current landscape for computational service providers in China is crowded, with over 600 companies already active in the sectorThis influx has resulted in heightened demands for resources, alongside an urgent call for upgrades and green transformations within the industryThe competitive dynamics are influenced not only by the burgeoning demand but also by external factors such as GPU (Graphics Processing Unit) supply policiesHigh Xiaoyue, chief researcher at the Global Leasing Industry Competitiveness Forum, posits that GPU servers will significantly impact the operational and economic dynamics of this marketThe shift towards domestic chip replacement is becoming a prevailing trend, urging industry players to evaluate the market potential for indigenous GPUs more proactively.

As evidenced by various governmental and financial reports, the demand for GPU technologies is anticipated to remain robust in the coming yearsNational Securities research indicates that over the next two to three years, GPUs will dominate the conversation surrounding domestic computational capabilities, reinforcing their dual virtue of being both usable and user-friendly.

On December 27, 2024, Lianhua Holdings announced its plans to enter a high-performance computational service agreement with a company in Shanghai

Advertisements

Their subsidiary, Lianhua Zixing Intelligent Computing Technology Co., is to provide 1,545 PFLOPS (Peta Floating Point Operations Per Second) of high-performance computational services, as part of a contract valued at 555 million yuanThis particular contract exemplifies the increasing focus on leveraging advanced computational technologies for commercial success.

Moreover, investment inquiries directed to Maxinlin revealed that a subsidiary, Suzhou Ruiying Intelligent Computing Technology Co., had signed contracts amounting to 320 million yuan with major corporations in September 2024. Such transactions reflect the escalating interest among listed companies to engage in computational leasing ventures.

In an interview, Xie Chuang Data shed light on its innovative approach wherein the company employs self-invested computational servers alongside third-party IDC (Internet Data Center) facilities to establish a capable computational service clusterWith service hubs already emerging in cities like Shanghai and Hangzhou, Xie Chuang aims to create a sustainable revenue stream through computational rental services.

However, several industry professionals have raised concerns about the significant upfront investments required to launch computational leasing operationsThe long return-on-investment periods, coupled with the rapid obsolescence of computational servers, present a notable risk for businesses venturing into this territoryConsequently, many enterprises are leveraging financial institutions and financing leases to secure operational capital.

In a disclosure by Lianhua Holdings, it was revealed that to ensure the fulfillment of its procurement plans, Lianhua Zixing must pay approximately 310 million yuan, with future projections indicating an equipment acquisition capital requirement of around 780 million yuan over the next two yearsThis scenario illustrates the pressing financing pressures faced by companies in the computational rental space.

Statistics suggest that the computational leasing market possesses significant potential

Advertisements

While Lianhua Zixing's proposed PFLOPS of service represents over 1,500 units, forecasts indicate that the global computational power may exceed 300 EFLOPS (Exa Floating Point Operations Per Second) by 2025, with intelligent computational services accounting for 35% of this figureThis forecast highlights the increasing reliance on computational resources driven by large model training and inference.

As seen from the rise of artificial intelligence, numerous financial leasing companies are keen on reshaping their business models to capitalize on the growing demand for GPU leasingOn September 25, 2024, Kingsoft Cloud publicized its agreement with a financial leasing firm to secure up to 300 million yuan for expanding its AI infrastructure investmentSimilarly, in October 2024, Qingyun Technology announced a financing deal through a sale-leaseback arrangement for GPU servers, underlining the fusion of GPU leasing into their business strategies.

This growth trajectory for AI models promotes an exponential increase in intelligent computational requirementsIn light of this explosive demand, several finance leasing companies are eager to pivot towards hardware rental operations, seeking to empower new productivity paradigmsHigh Xiaoyue suggests that as the scale of computational construction expands, investment costs will significantly increase, and thus securing financing will be integralVarious leasing companies are currently assessing potential involvement in this lucrative market, recognizing that operational dynamics of computational leasing differ substantially from traditional IDC approaches.

In the broader context of third-party intelligent computing centers, cost and revenue considerations play a crucial roleProcurement costs aside, ongoing operations involve depreciation, cabinet expenses, maintenance costs, and potential SLA-triggered penaltiesThe capacity for high-performance servers to attract substantial revenue hinges on maintaining a consistent high utilization rate, illustrating the fine balance between capital investment and revenue generation.

According to Zhang Xueli, founder of Shengqi Ruan, the computational market is undergoing rapid growth with surging demand across the board

Advertisements

Advertisements

Advertisements