Huawei Cloud Stack Pivots to Agentic AI Infrastructure

Antonony Gu, President of Huawei Hybrid Cloud, outlined three primary shifts during the keynote: the emergence of AI agents as a new class of digital employees, the move toward integrated, task-oriented resource delivery, and the consolidation of IT management around AI-specific assets. To support these changes, the company launched an AI agent-ready hybrid cloud architecture. This framework features an agile compute infrastructure, a capability hub for data and tokens, and an enablement platform for large-scale agent development.

Technological refinements include the deployment of A5 SuperPoDs, which utilize NPU pooling and memory snapshots to boost compute utilization from 30% to 70%. Beyond infrastructure, the company is focusing on the entire data lifecycle—from storage to intelligent decision-making—to ensure that enterprise data translates directly into productivity. The strategy extends to industry-specific foundation models, including vision-language and scientific computing tools, integrated into the Cloud Stack to facilitate real-world deployment.

Early adoption is already underway, with major financial and national infrastructure projects leading the transition. Standard Bank is leveraging the architecture to modernize its core banking systems across 20 countries, citing the need for zero-data-loss reliability. Meanwhile, Pakistan’s Sky47 has utilized the platform to build a national AI cloud, supporting generative AI applications across the finance, manufacturing, and energy sectors. By prioritizing these agentic capabilities, Huawei aims to bridge the gap between experimental AI development and production-grade industrial application.