智通财经网·2025-08-25 05:59Core Insights - The report from Guosen Securities highlights the sustained demand for AI computing power, with GPU cloud (computing power leasing) emerging as a key solution for the mismatch between high-end computing supply and demand in the short term, and a flexible, low-cost solution with increasing penetration in the long term [1] Industry Overview - The global computing power market continues to thrive, driven by the accelerated shipment of NVIDIA's GB series high-density computing cabinets, further enhancing the high-end computing landscape [1] - IDC forecasts that the global AI server market will reach $222.7 billion by 2028, with the share of generative AI servers increasing from 29.6% in 2025 to 37.7% in 2028 [1] Demand and Supply Dynamics - On the demand side, the need for training remains high due to rapid model iterations, while inference demand is gradually increasing as applications penetrate further [1] - On the supply side, high-performance computing chips represented by NVIDIA's B/Rubin and AMD MI series are continuously iterating, with the GB300 expected to accelerate delivery in the second half of 2025 [1] Capital Expenditure Trends - Major cloud service providers (CSPs) are expected to increase their capital expenditures significantly, with Amazon, Google, Microsoft, and Meta projected to collectively reach $361 billion in Capex by 2025, a year-on-year increase of over 58% [1] - Domestic companies like ByteDance, Tencent, and Alibaba are anticipated to exceed 360 billion yuan in Capex, indicating a sustained high demand for AI development [1] GPU Cloud Market Development - The GPU cloud (computing power leasing) model is rapidly evolving to address the global shortage of high-end AI chips, with major AI companies like Meta, Microsoft, and OpenAI announcing or completing the construction of large-scale clusters [2] - The GPU cloud market is expected to grow to $12.8 billion by 2033, marking a significant shift in the cloud computing landscape towards AI intelligent computing [2] Competitive Landscape - NVIDIA is supporting the development of GPU cloud providers (CoreWeave, NBIS, Omniva) through equity or partnership, solidifying its global leadership in high-end chips [3] - Domestic GPU cloud companies are positioned for growth, with rental returns estimated at 15%, reflecting a favorable comparison to overseas GPU cloud business models [3]