The commercialization of AI agents is accelerating, with downstream application demand being released in a concentrated manner, directly driving an explosive increase in computing power demand. This has significantly boosted the popularity of computing power rental services.
On March 12, the computing power rental concept sector performed actively. By the market close, shares of Hongjing Technology Co., Ltd. and China Energy Engineering Corporation Limited, among other related stocks, had hit the daily limit increase. Companies such as Beijing Geling Interactive Information Technology Co., Ltd. and Beijing HCR慧辰资讯 Information Technology Co., Ltd. also saw gains.
Computing power rental is expected to become a key solution for alleviating the current AI computing power supply gap. It can quickly meet the short-term computing power needs of AI enterprises, especially for small and medium-sized companies lacking sufficient funds and technical capability to build their own data centers, as it substantially reduces upfront investment costs. Simultaneously, it enhances the overall utilization efficiency of computing power resources, avoiding waste.
Meanwhile, rising rental prices have become a key issue for the market's subsequent development. According to industry monitoring, a surge in AI computing power demand at the beginning of 2026 has driven the computing power rental market into a price increase cycle. By the end of February, rents for high-end GPUs like NVIDIA's H200 and H100 had increased by 15% to 30% month-on-month. The hourly rent for the H200 reached 7.5 to 8.0 RMB per card-hour, with monthly rents between 60,000 and 66,000 RMB, marking a 25% to 30% increase. The monthly rent for the H100 rose to 55,000 to 60,000 RMB, a 15% to 20% increase. Delivery timelines have extended to the second quarter of 2027 for the H200 and the first quarter of 2027 for the H100.
In response to this trend, from an operational perspective, companies can hedge against the uncertainty caused by short-term price fluctuations by securing long-term contracts, such as fixed-term rental agreements for 3 to 5 years. From a long-term industry development perspective, it is necessary to accelerate the improvement of industry standard systems, clarify core indicators like computing performance and security levels, and guide orderly competition among market participants. Relevant policy subsidies should be precisely targeted towards small and medium-sized enterprises and research institutions in the sector to prevent resources from becoming overly concentrated in leading companies.
Amid the industry trends of rising prices and intensified competition, computing power rental companies should avoid恶性 price wars and instead build differentiated barriers by improving service quality and technical capabilities. Furthermore, these companies need to consider the cost pressures faced by traditional industries undergoing AI transformation. Measures such as launching special discounted computing power packages for traditional industries and collaborating with them to co-develop application scenarios can promote a synergistic development pattern between the computing power rental industry and the real economy, rather than merely seeking short-term gains through price hikes.
The high growth prospects of the computing power rental market are also attracting listed companies to accelerate their expansion into related businesses through various means. Recently, on investor interaction platforms, the progress of listed companies' computing power rental operations has become a hot topic among investors.
Leveraging its own computing power applications, HCR慧辰资讯 Information Technology Co., Ltd. has launched an all-in-one product for intelligent computing resource operation management called the "Integrated Computing Power Management Service Platform." This platform supports the integrated management of heterogeneous computing resources and provides services such as computing resource trading and operational management, better meeting the needs of various AI computing power rental scenarios. The platform has already been deployed and applied within the company and in relevant data centers. Regarding computing hardware, a small amount of computing resources have also been leased to related clients.
Hunan ABL Environmental Protection Technology Co., Ltd. announced that its computing power center's rental business commenced in October 2024, with the overall construction completed in December 2024. Since the center's completion, the company has gradually strengthened its computing power rental business capabilities, enhanced team building and technical reserves, and further improved its market competitiveness. Consequently, the company's sales and profits from computing power rental business in 2025 showed significant growth compared to 2024.
Sharewin Data Technology Co., Ltd. stated on an investor interaction platform that computing power services have become an important business direction for the company, which has already signed cooperation or service agreements with multiple clients.
Looking ahead, analysts believe that companies possessing technologies for computing-power synergy,布局 in green power computing, and strong customer resources will hold a dominant position in the computing power rental market competition and become core beneficiaries of AI industry development. In the long run, the development of computing power rental still needs to address bottlenecks such as technological dependency and security compliance. The industry is expected to evolve towards greener practices, service tiering, and ecological collaboration. Companies will need to achieve sustainable development through hybrid rental models, technological cost reduction, and policy coordination.