China Securities Construction Investment reports that domestic AI models are accelerating iteration, with GLM-5.2 and Kimi K2.7 strengthening 1M context, long-range Agent, and Agentic Coding capabilities. This evolution pushes domestic models from general Q&A toward developer tools and enterprise-level workflows.

Meanwhile, Kimi enhances international operations, DeepSeek's financing boosts industrialization expectations, and WeChat's AI gray-scale testing indicates AI entry points are moving from independent apps to super app ecosystems, likely increasing model call frequency and inference token consumption.

Overseas, Anthropic sees upward revisions in revenue, major clients, and financing scale, signaling continued expansion in enterprise workflow and Agentic Coding demand. On the infrastructure side, AI cloud vendors face growing backlogs while GPU rental prices rise. Supply constraints for CPU and DRAM/NAND are spreading, suggesting the boom in computing power, storage, and rental services will persist.