Huawei announced a broad AI infrastructure portfolio during its ID 2026 forum in Paris, unveiling new AI data storage, pipeline acceleration, and agent development technologies designed to support enterprise-scale artificial intelligence workloads.
The company introduced what it calls an “AI Data Platform,” a full-stack ecosystem that combines AI training storage, inference acceleration, data ingestion systems, and AI agent runtime frameworks. Although Huawei did not directly use the term “AI factory,” the platform closely resembles similar AI infrastructure strategies from global technology vendors.
Huawei’s new ecosystem includes OceanStor Pacific data lake storage, OceanStor Dorado all-flash systems, OceanStor A800 AI storage appliances, and Omni-Dataverse ingestion tools. The company also integrated backup and archive support through its OceanProtect platform.
Yuan Yuan said the future of AI will increasingly depend on data infrastructure innovation. Huawei aims to help enterprises accelerate AI adoption through advanced storage technologies and integrated AI systems.
Huawei Builds Full Stack AI Infrastructure Platform
Huawei’s AI platform combines high-capacity storage systems with AI-specific acceleration technologies to support large-scale model training and inference tasks. The architecture is designed to manage massive datasets, vector databases, and AI pipelines efficiently.
The OceanStor Pacific platform serves as Huawei’s scale-out AI data lake storage layer. The system can reportedly provide up to 11 petabytes of storage capacity in a compact 2RU chassis using Huawei’s high-density SSD technology.
Huawei also introduced the DME Omni-Dataverse platform, which enables real-time ingestion and management of multimodal enterprise data across multiple locations. The system supports global visibility, unified management, and large-scale vector retrieval for AI applications.
According to Huawei, Omni-Dataverse can retrieve hundreds of billions of high-dimensional vectors within seconds, helping organizations support retrieval-augmented generation (RAG), AI search, and intelligent agent workflows.
Huawei Expands AI Storage and Data Acceleration Technologies
The A800 storage system includes data processing units (DPUs) and GPU Direct-like functionality to accelerate AI workloads. Huawei says the platform supports fast checkpointing, high-bandwidth training set loading, and efficient handling of vectors, tensors, and RAG data pipelines.
Huawei also introduced a specialized Data Engine Node that processes AI-specific data workloads on top of Dorado storage systems. The node can generate vector embeddings, build knowledge bases, and manage memory structures required for advanced AI applications.
The company positions the Data Engine Node as a way to make general-purpose storage systems more AI-ready without requiring organizations to completely rebuild existing infrastructure environments.
Read : Litter-Robot CMO Uses Marketing to Boost Cat Ownership
Unified Cache and AI Agent Systems Support Enterprise AI
Huawei’s architecture also includes a Unified Cache Manager (UCM) layer designed to accelerate data access across its AI storage ecosystem. The system works alongside knowledge bases, memory banks, and KV cache stores to optimize AI model operations.
The UCM acts as part of Huawei’s broader Context Memory System (CMS), which is intended to improve how AI models and agents retrieve, manage, and utilize contextual information during inference and decision-making tasks.
In addition to storage infrastructure, Huawei also unveiled AI agent development and runtime frameworks. These tools aim to support the creation, deployment, and operational management of enterprise AI agents and autonomous workflows.
As geopolitical tensions continue limiting Huawei’s access to U.S.-based GPU suppliers such as NVIDIA, the company is increasingly building its own AI ecosystem around internally developed Ascend processors, Atlas accelerators, and proprietary AI software frameworks.
