Huawei Cloud has announced a suite of enterprise AI products aimed at supporting “agentic” AI, in which software agents can operate autonomously in production environments. The rollout includes new infrastructure, model access and enterprise agent tools, reflecting Huawei’s effort to build an open cloud ecosystem combining computing, memory, models and industry applications.

The latest AI Cluster Service is claimed to support more than 40 days of stable training, recover faults within 10 minutes and deliver 20% higher token throughput than its predecessor; it is due in China on 30 September and elsewhere on 30 November. Huawei also described petabyte-scale Context Memory Storage, an Agentic Model as a Service platform offering one-click access to external models, and AgentArts alongside its open-source edition, openJiuwen. The company says its agent infrastructure serves over 3,500 customers, while its enterprise platform supports more than 100 organisations and provides over 5,000 general-purpose and 1,000 industry-specific Model Context Protocol assets.