GLMS Securities has released a research report stating that domestic daily API call volumes have surged over a thousandfold in two years. This supply-demand imbalance is driving sustained price increases for computing power leasing and cloud services both domestically and internationally, while the implementation of tiered consumer payments validates the commercial value of Tokens. Top-level institutional design has established a closed-loop chain from data supply to model development and finally to real-world application deployment. This framework solidifies the core role of the Token as a value carrier, providing a foundation of certainty for industry development. Token "inflation" represents the central theme of the current AI industry market movement, with non-linear demand growth driving a revaluation across the entire industry chain encompassing computing power, large language models, and applications. The report highlights leading companies with clear technological moats and proven commercial models as key beneficiaries.
GLMS SEC's primary viewpoints are outlined below: Under the overarching trend of Token "inflation," the AI computing power industry is entering a new growth cycle driven by converging factors of demand, architectural shifts, and domestic substitution.
Token "Inflation": Domestic daily API call volumes have surged over a thousandfold in two years. The resulting supply-demand gap is pushing up prices for computing power leasing and cloud services globally. Concurrently, the rollout of tiered consumer payment structures validates the commercial value proposition of Tokens.
Industry Bottleneck Shift: The core industry challenge is transitioning from an overall shortage of GPUs to unlocking system-level efficiency. In Agent scenarios, high CPU processing latency proportion is the primary bottleneck limiting GPU performance utilization. The CPU has evolved from a traditional supporting component to a core variable in AI infrastructure. Efficiency-enhancing technologies like hyper-nodes and compute-storage optimization are being deployed in parallel.
Historic Opportunity for Domestic Computing Power: The deep integration of large language models with domestic computing solutions is breaking dependence on foreign technological pathways. Core manufacturers are poised to maintain high-speed earnings growth. Furthermore, the Token operational model upgrades the transaction from mere computing power sales to intelligent service provision. This business model reconstruction opens up substantial long-term growth potential.
The large language model industry is entering a triple-driven growth phase characterized by business model validation, rapidly expanding demand, and the implementation of domestic substitution.
Overseas Benchmark for High-Level Intelligence: Companies like Anthropic have demonstrated a viable commercial path for advanced AI, achieving a 10x annualized growth in Annual Recurring Revenue (ARR) and valuations surpassing the trillion-dollar mark. This validates the industry logic that "the upper limit of model intelligence directly dictates pricing power." The trend of higher-tier models achieving both volume and price increases is clear.
AI Agents Fuel Token Demand: The primary consumers are shifting from humans to machines, removing the demand ceiling previously constrained by human population size. Resident intelligent agents, exemplified by projects like OpenClaw, are evolving model interactions from single-turn Q&A to long-chain, multi-tool execution. The industry's value metric is switching to Tokens. Telecom operators and major internet firms are launching tiered service packages, upgrading the business model from computing power transactions to intelligent service exchanges.
Rising Importance of Sovereign AI: Restrictions on exporting cutting-edge models underscore the critical importance of sovereign AI capabilities. Domestic large language models have entered a new development stage marked by localized breakthroughs and scaled validation.
The AI application industry is maturing rapidly along the paths of institutional framework development, vertical industry penetration, and business model innovation.
Institutional Foundation: Top-level design has built a full-chain closed loop from data supply to model development and application deployment, cementing the Token's role as the core value carrier and providing stable support for industry growth.
Vertical Industry Progress: Large models for specific sectors have moved from technical proof-of-concept to establishing commercial closed loops. Collaboration models are shifting from traditional project-based contracts to value-sharing arrangements. Leading firms have pioneered subscription-based business models integrated with Token-based pricing, creating a clear replication path.
Native Business Models: AI Agent-driven native commercial forms have completed initial large-scale validation in specific scenarios. Payment use cases have successfully demonstrated a full closed-loop process from intelligent agent capability invocation to transaction completion, establishing a foundational trust mechanism for the broader intelligent agent economy.
Regarding investment targets, the report recommends leading companies such as Knowledge Atlas (02513), Cambricon (688256.SH), Inspur Information (000977.SZ), Hygon Information (688041.SH), and Supcon Technology (688777.SH).
Risk warnings include the potential for slower-than-expected technological iteration, delays in commercial implementation, and intensifying industry competition.
Comments