Huawei Cloud announced a global expansion of its enterprise AI product line at HUAWEI CONNECT 2026 in Shanghai on September 18, rolling out new infrastructure, an agent development platform, and two industry-specific zones designed to move agentic AI from pilot projects into production. The moves target enterprises that need to run AI agents securely at scale, not just experiment with models.
Dr. Peter Zhou, Director of the Board at Huawei and CEO of Huawei Cloud, said the company's infrastructure strategy has shifted. "In the agentic era, infrastructure no longer simply exists to provide compute. Instead, its purpose has expanded to making every token more efficient, coordinating general and AI compute, enabling models to keep learning, and, more importantly, running agents securely and reliably within real production environments."
Infrastructure built for agent workloads
The latest AI Cluster Service (AICS) launches commercially in China on September 30 and in international markets on November 30. It includes a five-level fast recovery mechanism with full-chain observability, supports more than 40 days of stable training on cloud, and recovers from faults within 10 minutes. Coordinated optimization across scheduling, cache, and algorithms delivers 20% higher token throughput than the previous generation.
Huawei Cloud also introduced Context Memory Storage (CMS), a petabyte-scale memory solution for long-horizon agent tasks. It offers twice the storage capacity of comparable products and supports terabyte-scale memory reads at speeds the company says are 50% higher than industry peers. The Agentic Model as a Service (MaaS) platform aggregates models from multiple providers, letting developers call state-of-the-art models with one click and no deployment overhead. MiniMax demonstrated its multimodal model on the platform at the event.
An enterprise agent platform with open-source backing
The company's agent platform runs on a dual commercial and open-source track. AgentArts, the commercial edition, and openJiuwen, the open-source version, together expose more than 5,000 general Model Context Protocol (MCP) assets and over 1,000 industry-specific MCP assets. The platform already serves more than 100 enterprises, including China Southern Power Grid, Kingsoft Office, and the Shenzhen Longgang District Government.
Kingsoft Office shared its deployment at the event: the company integrated the agent platform with WPS 365 Document Center and WPS Comate to build vertical office agents now running across finance and government sectors. AgentArts will be commercially available outside China on December 30. The openJiuwen community has passed 50,000 stars and 3.29 million downloads. Huawei Cloud also launched partner commercial editions with Chinasoft International, iSoftStone Group, and Beiming Software. For teams exploring how to build and deploy agents in production, understanding the broader category of AI Agents & Automation is becoming essential.
Industry AI Foundry adds government and hardware zones
The Industry AI Foundry now holds more than 1,000 industry assets and has supported over 1,000 deployed projects across five existing zones: Smart Healthcare, Embodied AI, AI for Science, Smart Manufacturing, and Smart Finance. Two new zones launched at the conference. The Smart Government Zone has 24 founding partners covering government office operations, public services, urban governance, and industry development. The AI Hardware Zone has onboarded 15 core partners spanning more than 20 device types - including AI glasses, AI toys, and AI recording cards - plus over 10 scenario templates and 110 scenario skills.
These additions signal a push to turn isolated project experience into reusable, industry-specific capabilities. For IT and development leaders evaluating infrastructure and platform decisions, the implications extend into AI for IT & Development strategy and tooling choices.
Why this matters for executives and IT leaders
Huawei Cloud is making a concrete bet that the next phase of enterprise AI is not about models alone - it is about the infrastructure, memory, and platform services required to run agents in production. The AICS recovery guarantees, the CMS memory specs, and the MaaS multi-model access are all designed for teams that have moved past prototyping. For executives, the timeline is clear: AgentArts reaches international markets in December, and the infrastructure is available within weeks. The question is whether your organization's agent workloads are ready for it.
Your membership also unlocks:
