Logo

Enhancing intelligence: Huawei's full-stack data infrastructure for AI centres

Huawei has introduced a data infrastructure aimed at supporting AI-driven enterprise environments, covering areas including AI data lakes, model engineering, and data resilience.

The Huawei IDI Forum 2026 included a presentation by Yuan Yuan, Vice President of Huawei and President of the Data Storage Product Line. The event featured the introduction of Huawei’s full-stack data infrastructure intended for AI Data Centres (DCs). The initiative is positioned as supporting the development of industry intelligence and the digital transformation of enterprises.



Enterprises are currently experiencing increased adoption of AI-driven systems. Digital agents are being used in productivity workflows and, in some cases, are evolving into digital roles within organisations. This trend is contributing to changes in traditional IT architectures to better accommodate AI-focused requirements. Huawei’s approach includes several components: data lakes, AI data platforms, compute capabilities, models, agent frameworks, and data resilience mechanisms.

Within the AI data lake area, Huawei references OceanStor Pacific Scale-Out Storage, which provides high-density storage designed to optimise storage costs. The DME Omni-Dataverse is also included, supporting multiple data types and enabling real-time data ingestion.

In the area of knowledge and memory platforms, Huawei introduced Context Memory Storage (CMS), which is designed to support heterogeneous computing and reduce token retrieval times in data processing workflows.

For AI inference workloads, a 3+1 AI data platform is described that combines KV cache acceleration with a knowledge base and reports retrieval accuracy of over 95%. The Unified Cache Manager (UCM) is included to improve inference performance through scheduling optimisation.

For model engineering and resource management, Huawei’s ModelEngine is presented as a platform that supports model deployment and compute resource partitioning for scheduling and utilisation purposes.

The agent framework and data resilience components are described as providing end-to-end support for AI data protection. The ModelEngine Nexent platform includes functions that can adjust agent skills based on natural language input.

Overall, the forum presentation outlined Huawei’s data infrastructure components related to AI-oriented systems and their intended use in enterprise environments.
Ben Wilkes joins Portman Partners, enhancing executive search capabilities during a significant...
New sovereign UK inference cloud is made possible by energy-efficient SambaNova AI infrastructure.
AMD GPUs are now supported on Dell PowerEdge servers, extending on-premises AI compute capabilities.
Dell Technologies has introduced PowerMaxOS 10.4, with updates focused on performance improvements,...
Yan Evans has been appointed Managing Director for the Global Data Centre Business at Güntner...
BT Business has launched a sovereign services portfolio designed to support organisations with...
Kao Data partners with Discover Tech, aiming to provide immersive tech sector experiences for young...
A demonstration of hydrogen-fuelled engines has been completed as part of testing for data centre...