Tencent Cloud announced that AI agents are currently generating over 20,000 instances of the Tencent Cloud TDSQL-C database daily. This figure demonstrates a growth of more than 80 times over the past year. Over one million such instances are already in active use.
A Tencent Cloud Vice President explained this transformation by stating that although the first user of the database has changed, its core function remains the same. When developing applications, an agent needs to quickly create a database and verify changes in an isolated environment. After launching the application, the database must scale with the load without disrupting business processes.
To ensure this functionality, TDSQL-C has updated its serverless scaling capabilities and added a branching feature. This feature allows the agent to test a new approach on a separate copy of the database before its release. During a stress test, the serverless version was able to scale from 0.25 to 32 compute units in 75 seconds and returned to 0.25 within 10 seconds after the load ended.
Furthermore, TDSQL-C is integrated with CloudBase, enabling the agent to create tables, set permissions, and generate APIs during application development. The PostgreSQL version of CloudBase is fully compatible with Supabase. The company Zhihu uses CloudBase to deploy applications created with AI coding tools, while the Japanese gaming company Mynet allows employees to ask questions about game data in natural language through separate read-only compute resources.
Tencent Cloud also introduced TDSQL Nexa Lakebase—a multimodal database that transforms table fields, business terms, and metric definitions into context accessible to agents. Additionally, TencentDB Agent Memory, which was released in May, now supports shared command memory as well as individual agent memory, boasting over 30,000 stars on GitHub.
The company Li Ning serves as an example of agent usage. Li Ning's General Manager of Information Systems told China Times that the sports brand implemented Tencent's WorkBuddy office agent in its supply chain and merchandise departments, launching approximately 50 AI agents in total. Together, they provide about 460,000 hours of equivalent work per year, with two-thirds of this volume attributed to the digital store manager agent. Li Ning plans to increase the number of agents to 200 within the year and reach 4.6 million equivalent hours next year.
Tencent noted that since the launch of WorkBuddy in March, it has executed 9 billion task requests. Dawson Tong, head of Tencent's Cloud and Intelligent Industries Group, emphasized that the key to deploying agents is connecting people, tools, systems, and data.