From "devouring" to "benefiting," Chinese software is expected to enjoy "lobster meat"!
```
Once suppressed by the narrative of “AI devouring SaaS,” the software sector is now undergoing a new wave of value re-evaluation: the explosion of AI Agent-driven Token usage is shifting the core value of software from “function presentation” to “process orchestration, data and permission management,” with infrastructure and platform-type software likely to be the first to capture the incremental gains.
The latest catalyst from the industrial side comes from the breakout success of OpenClaw. According to HSBC Qianhai, the Token call volume of AI Agents can reach up to 15 times that of chatting; on the OpenRouter platform, the weekly Token call volume rose to 14.8 quadrillion in the week of March 2, increasing around 160% in two months, pushing up curves for both AI application monetization and computing power consumption.
For the A-share market, HSBC Qianhai Securities believes that strong growth in Token call volume will boost demand for AI software infrastructure and IDC, and Chinese models are continuously increasing their global Token share thanks to higher cost-efficiency.
AI Agents: From “Chatting” to “Executing,” Amplifying Token Consumption
HSBC Qianhai Securities believes that as foundation models improve their capabilities in productivity tasks, AI Agents are becoming the “first blockbuster AI application.” OpenClaw is a representative example, capable of automatically completing software installation, script execution, file organization, and website monitoring tasks.
Popularity and usage are rising in tandem. HSBC Qianhai Securities reports that since its launch in November 2025, OpenClaw’s stars on GitHub have exceeded 299,000—more than well-known open-source projects like Linux and React.
More critically, the consumption structure is changing: according to Anthropic data, Token consumption by AI Agents can reach up to 15 times that of chat interactions, making Token volume a “hard metric” for measuring the penetration of AI applications.
OpenRouter’s data reinforces this trend. HSBC Qianhai Securities points out that, driven by individual users and enterprises actively deploying multi-agent systems like OpenClaw, weekly Token calls on the OpenRouter platform have grown rapidly since early 2026 and hit 14.8 quadrillion in the week of March 2.
Chinese Models Seize Share: Cost Advantages Amplify Infrastructure and IDC Flexibility
HSBC Qianhai Securities points out that since February 2026, Chinese models such as Zhipu GLM5, Minimax M2.5, and Moonshot Kimi K2.5 have been steadily capturing U.S. large model Token market share with their performance and cost-efficiency advantages.
According to OpenRouter data, in the week of February 9, for the first time, Chinese models accounted for a greater proportion of weekly Token calls among the top nine models on OpenRouter than U.S. models.
The pricing structure is also driving penetration. Summary of major AI programming service subscription packages by HSBC Qianhai Securities shows that with similar usage caps, Chinese vendors’ monthly subscription prices are only about 20% to 30% of overseas vendors’, making cost-efficiency a key factor in Token migration.
HSBC Qianhai Securities judges that strong growth in Token call volume will increase demand for AI software infrastructure, with Hyper-Converged Infrastructure (HCI) and Software-Defined Storage (SDS) likely to see structural opportunities; meanwhile, the increasing adoption of AI Agents is expected to drive order growth for domestic IDC providers.
OpenClaw Goes Viral: Agents Are Not Software Alternatives But Act as an “Invocation Layer” Above
Shenwan Hongyuan Research notes that OpenClaw is positioned as a personal AI assistant that can connect to multiple communication channels such as WhatsApp, Telegram, and Feishu. Through a local gateway it manages conversations, tools, events, and skills in a unified way, showing that Agents are moving towards an “executable” product form.
From an architecture perspective, OpenClaw uses a local WebSocket gateway for authentication, orchestration, configuration management, and log processing; applications and devices access as nodes and expose their command execution, messaging, and device capabilities to the Agent.
Shenwan Hongyuan concludes that Agents are not new entities divorced from software, but rather an “intelligent invocation layer” and “task orchestration layer” built upon existing software, data, and permissions structures, with their core value in improving automation and process management efficiency.
According to Shenwan Hongyuan Research, the pressure previously faced by the software sector mainly came from the AI rewriting valuation frameworks, not from a deterioration of fundamentals. At the start of the year, investors worried that general-purpose large models and general-purpose Agents might bypass traditional SaaS interfaces, directly eroding functional value and user stickiness, creating even an “end of SaaS” extreme narrative.
The marginal change is that enterprise-level Agents cannot take effect simply by “plugging in a model”, but must deeply integrate with enterprise data and permission controls, establish security protections, and achieve observability, evaluability, and auditability.
Software companies are not sitting ducks; on the contrary, they're leveraging their existing process entry points and data barriers to rapidly productize Agents. ServiceNow explicitly emphasizes that its AI Agents are built-in, placing the AI agent, data, and workflow on a single platform and unifying third-party Agents.
The software sector now looks more like it's in the early stages of recovery after sentiment has bottomed out: pessimistic narratives have not completely disappeared, but indiscriminate selling is over, and the market is beginning to assess who is least impacted or may even benefit from the AI wave.
Risk DisclaimerThe market involves risks, and investments should be made cautiously. This article does not constitute personal investment advice, nor does it take into account individual users’ specific investment goals, financial situations, or needs. Users should assess whether any opinions, views, or conclusions in this article fit their own circumstances. You are responsible for any investment made based on this article. ```