Guolian Minsheng Securities: Token "Inflation" is the core theme of this round of AI industry market.

Guolian Minsheng Securities released a research report stating that the daily call volume in China has increased by over a thousand times in the past two years, leading to a supply-demand gap in both domestic and overseas computing power rental and cloud service prices continue to rise. At the same time, the landing of C-end hierarchical payment validates the commercial value of Token. The top-level institutional design has built a complete closed loop from data supply to model development to scenario landing, clearly defining Token's core positioning as a value carrier, providing deterministic support for industrial development. Token "inflation" is the core theme of this round of AI industry trends, with non-linear growth on the demand side driving a revaluation of the entire industry chain of computing power, large models, and applications, with a focus on leading companies with clear technological barriers and thorough commercial verification. Guolian Minsheng Securities' main points are as follows: Under the trend of Token "inflation," the AI computing power industry is ushering in a new growth cycle of demand, architecture, and domestic substitution resonance. Industry contradictions have shifted from a shortage of total GPU to the release of system-level efficiency: In the Agent scenario, the high percentage of CPU processing delay is the core bottleneck restricting the release of GPU efficiency. CPUs have jumped from traditional supporting roles to core variables in AI infrastructure, and efficiency improvement technologies such as super nodes and storage calculation optimization are being implemented simultaneously. Domestic computing power is entering a historic development window of model-algorithm synergy: Deep integration of large models and domestic computing power breaks the dependence on overseas technological paths. Core manufacturers are expected to maintain high-speed growth in performance, while Token operational models upgrade computing power trading to intelligent service trading, restructuring business models and opening up long-term growth spaces. The large model industry is entering a stage of triple-driven growth characterized by business model validation, rapid demand growth, and domestic substitution. Overseas benchmarks have explored commercialization paths for high-level intelligence: Anthropic achieved annualized growth of 10 times ARR, with a corresponding valuation exceeding one trillion USD, verifying the industry logic that "the upper limit of model intelligence directly determines pricing power," with a clear trend of simultaneous rise in quality and quantity of high-level models. AI Agent drives the growth of Token demand: The consumption subject has shifted from humans to machines, and the demand limit is no longer constrained by population size. Persistent intelligent agents like OpenClaw have extended model calls from single-round Q&A to long-chain multi-tool execution, with industry value measured in Tokens. Operators and internet giants are introducing tiered packages, transforming the business model from computing power trading to intelligent service trading. The importance of sovereign AI is highlighted: Restrictions on exporting advanced models highlight the importance of sovereign AI. Domestic large models have entered a new stage of partial breakthroughs and scaled verification. The AI application industry is accelerating maturity along the paths of institutional design, vertical landing, and business model innovation. The top-level institutional design builds a complete closed loop from data supply to model development to scenario landing, clearly defining Token as the core positioning of value carrier, providing deterministic support for industrial development. Large models in vertical industries have transitioned from technical validation to commercial closed-loop stage, with cooperation models shifting from traditional project-based to value-sharing, with leading companies pioneering subscription-based business models combined with Token pricing, clear replication paths. The original business form driven by AI Agent has already completed the scale validation of the first scenario, with the payment scenario running from intelligent agent capability invocation to transaction completion, establishing a basic trust mechanism for the comprehensive development of the intelligent agent economy. Regarding the target: Recommended leading companies include KNOWLEDGE ATLAS (02513), Cambricon (688256.SH), Inspur Electronic Information Industry (000977.SZ), Hygon Information Technology (688041.SH), Zhejiang Supcon Technology Co., Ltd. (688777.SH) and others. Risk warning: Risks of unexpected technological iteration; risks of unexpected commercialization; risks of intensified industry competition.