AI专题：AI模型迭代聚焦工程能力，AI应用落地锚定高ROI场景

Investment Rating - The report indicates a shift in AI investment from FOMO CapEx to ROI CapEx, suggesting a more cautious and strategic approach to capital expenditures in the AI sector [4]. Core Insights - Overseas AI investments are facing cash flow pressures, prompting tech giants to seek various data center construction methods and financing solutions to alleviate these pressures. The IPO process for AI unicorns is expected to accelerate [4]. - The focus on maximizing token output efficiency per watt in data centers is becoming critical due to power capacity limitations. Companies are optimizing hardware and software to enhance computational efficiency [4]. - The engineering capabilities of large AI models are improving, with a stronger emphasis on commercializing AI products through various business models such as subscriptions, APIs, and advertising [4]. - The growth of AI cloud services is anticipated to accelerate as capacity is released, with significant orders expected in 2025 [4]. Capital Expenditure Rhythm - Capital expenditure expectations are being raised, with cash flow pressures increasing in the future [5]. - The AI infrastructure paradigm is shifting, with capital expenditures continuing to expand as companies transition from CPU to GPU-based workloads [10][12]. Data Center Construction - Data centers are facing power capacity limitations, leading to a focus on maximizing output efficiency per watt [6]. - The construction of data centers is increasingly flexible to accommodate various generations of GPUs and other components [4]. Model Capability Evolution - The evolution of AI models is expected to continue, with advancements in long text processing, multi-modal capabilities, and logical reasoning [7]. - The commercial viability of AI products is expected to increase as engineering capabilities improve [4]. AI Cloud Business Growth - The AI cloud service sector is entering an expansion phase, with significant orders expected and a rapid increase in contract values [8]. - The release of computational capacity is projected to drive accelerated growth in AI cloud services [4]. Capital Expenditure Cash Flow Statement - The capital expenditure of major tech companies is growing rapidly, leading to increased pressure on free cash flow [20][21]. - The ratio of capital expenditure to operating cash flow is at historically high levels, indicating sustained investment despite cash flow pressures [21]. Capital Expenditure Balance Sheet - The fixed asset scale of major tech companies is steadily increasing, with operating lease liabilities showing slight growth [26]. - The ratio of operating lease assets to fixed assets indicates a preference for self-built data centers over leasing [26]. Capital Expenditure Income Statement - Depreciation and amortization expenses are increasing, putting pressure on profit margins [30]. - The proportion of depreciation to revenue is at its highest level since early 2020, indicating rising costs associated with infrastructure investments [31]. Capital Expenditure Financing Needs - Tech companies are increasingly turning to debt financing to manage cash flow pressures resulting from high capital expenditures [36]. - Major tech firms have issued significant amounts of debt to support AI infrastructure investments, indicating a shift from cash investments to debt financing [36][38].