Huawei Shifts Cloud Focus to AI Agent Infrastructure

Huawei has announced a major pivot in its cloud infrastructure, positioning its hybrid cloud stack specifically for the era of autonomous AI agents. At a recent summit in Shanghai, the company outlined a new architectural approach designed to handle the complex demands of machine-based workers rather than just human users. This shift marks a move away from traditional IT management toward a system that prioritizes the continuous interaction between AI models and enterprise data.

The core of this strategy is a deterministic hybrid cloud architecture. According to Antonony Gu, President of Huawei Hybrid Cloud, the business landscape is changing as AI agents become digital employees. These agents require compute, data, and models to be packaged and readily available for immediate use. The new architecture aims to provide this integrated delivery, moving away from layer-by-layer assembly to a more seamless, task-oriented resource provision.

Redefining Resource Management for Agents

Traditional cloud management focuses on static resources like storage and networking. The new Huawei approach shifts the focus to managing AI compute, data, and models as unified entities. This change is critical because managing these components through separate systems often creates new silos, which can hinder the speed and efficiency required for agentic workflows. By integrating these elements into a single platform, the company aims to reduce the complexity of deploying and scaling AI agents.

The architecture includes an AI-native capability hub that connects data, compute, and tokens. This hub acts as a central nervous system for the AI ecosystem, allowing agents to access the necessary resources without manual intervention. The goal is to make resource provisioning agile and elastic, supporting continuous evolution and unified general-purpose and AI compute. This integration is designed to streamline the development and execution of AI agents at scale.

Infrastructure Upgrades for Higher Efficiency

To support these demands, Huawei is updating its physical infrastructure with technologies like NPU pooling and memory snapshots. These improvements are claimed to increase compute utilization from 30% to 70%, a significant boost for resource-intensive AI tasks. The infrastructure also provides an observable and governable framework for AI agents, turning standard cloud management platforms into intelligent hubs. This ensures that the underlying hardware is not only powerful but also manageable and secure.

Security is a key component of this evolution. The new stack offers end-to-end protection for infrastructure, models, and AI agents throughout their lifecycle. This comprehensive security model is designed to address the unique risks associated with autonomous systems, ensuring that data integrity and system stability are maintained. The trade-off is that enterprises must adopt a new management paradigm, which requires shifting from traditional IT oversight to a more dynamic, AI-centric operational model.

Transforming Data into Productivity

Huawei emphasizes the role of data in driving intelligent productivity. The cloud stack provides end-to-end data lifecycle capabilities, including reliable management, efficient delivery, and trusted circulation. By using a decoupled storage and compute architecture, the system can handle multimodal data more effectively. Tools like AI DataLake and DataArts are integrated to help enterprises turn raw data into actionable insights, laying a solid foundation for the agentic AI era.

The company is also investing in foundation models and integrating them with industry-specific knowledge. This includes domain-specific models for vision, scientific computing, and prediction, as well as optimization solvers. By bringing AI into production across various industries, Huawei aims to automate model iteration and continuous evolution. The source, GN auto tech/cloud, notes that this approach requires enterprises to rethink their data strategies, focusing on quality and accessibility to fully leverage the potential of AI agents.