Pacific Securities: OpenClaw expands AI use cases, driving the AI industry chain from infrastructure to model side.

Pacific Securities released a research report stating that the number of stars on GitHub's OpenClaw platform quickly surpassed 250,000, making it one of the fastest-growing open source projects on the platform. The token consumption of AI agents has significantly increased compared to traditional AI conversations, and domestic models have stood out due to their cost-effectiveness. Domestic technology giants actively launched Openclaw related products and services, and various local governments have also introduced supportive policies. OpenClaw continues to break barriers, further driving the AI industry chain from infrastructure to model end. Pacific Securities' main points are as follows: OpenClaw promotes AI from dialogue suggestions to autonomous execution OpenClaw was created by Austrian developer Peter Steinberger in November 2025, originally known as Clawdbot, and after two name changes, it was finally named OpenClaw. As a locally prioritized open-source AIAgent framework, OpenClaw can directly operate computers through natural language commands and execute practical tasks 24/7, achieving a transition from "dialogue suggestions" to "automated execution". Since the beginning of the year, the project's number of stars on the GitHub platform has rapidly surpassed 250,000, making it one of the fastest-growing open source projects on the platform. OpenClaw drives the consumption of large model tokens to reach new highs, and domestic models stand out due to cost-effectiveness The token consumption of AI agents has significantly increased compared to traditional AI conversations. In Chatbot mode, a user's single query corresponds to a model's one response. However, in Agent mode, OpenClaw engages in multiple rounds of reasoning when executing tasks, and performing a single task may trigger thousands of model calls, resulting in token consumption for a single Agent user that may be tens to hundreds of times higher than that of traditional conversation users. OpenRouter data shows that the consumption of large models has reached new highs recently, with a steep growth rate. As of the week ending March 4th, OpenRouter's token usage had reached 14.8T, doubling since January. Chinese models such as Minimax M2.5, Deepseek V3.2, and Kimi K2.5 rank in the top ten in terms of token consumption. Domestic technology giants actively release Openclaw related products and services ByteDance's volcano engine launched the ArkClaw cloud-based SaaS version, allowing users to use it by opening a web page, and supporting plugins for platforms like Feishu; Alibaba Cloud provides one-click deployment services, introducing the open-source Team version OpenClawHiClaw and the personal intelligent body workstation Copaw, which can be deployed locally or in the cloud, and support remote operations on Feishu and DingTalk; Baidu Cloud launched the mobile version of OpenClaw and the Red Fingertip Operator, supporting cloud phone deployment. Tencent released the OpenClaw-like product WorkBuddy and integrated it with WeChat for Enterprises and QQ; Xiaomi officially launched the mobile end system-level intelligent body Xiaomimiclaw for small-scale closed testing on March 6th. Local governments introduce special support policies for Openclaw From March 6th to 9th, 2026, cities such as Hefei, Shenzhen, Wuxi, and Changshu introduced special policies to support the development of the openclaw ecosystem in multiple dimensions such as funding, computing power, data, and scenarios. For example: 1) Hefei provides up to 10 million yuan in funding support for open-source AI projects, and offers computing power vouchers, corpus vouchers, and other subsidies. 2) Longgang in Shenzhen encourages the establishment of a "Crayfish Service Area" to support companies in purchasing or building OpenClaw intelligent body solutions, with a maximum of 2 million yuan per enterprise per year. 3) Wuxi provides up to 5 million yuan in support, covering industrial large model development, computing power subsidies, data annotation, and other aspects. Risk reminder: AI technology development may not meet expectations, and market competition may intensify.