Huawei Launches AI Data Platform at MWC
At MWC Barcelona 2026, Huawei launched a new AI Data Platform. The company stated the platform is designed to bridge the gap between AI models and business value by integrating technologies for knowledge generation and management. The announcement was part of a broader push by Huawei into AI-focused networking and data infrastructure.
The platform's core architecture, presented by Huawei's President of Data Storage Yuan Yuan, integrates a knowledge base, KV cache, and memory bank under a Unified Cache Manager (UCM). This design aims to solve key AI inference challenges, targeting a multimodal knowledge retrieval accuracy of over 95% and reducing the time to first token by 90%. Deployment is offered in two distinct hardware configurations: a greenfield "appliance mode" built on the OceanStor A800 system and an "independent mode" that upgrades existing systems by adding AI data engine nodes to OceanStor Dorado storage. This hardware-software integration is a component of Huawei's broader AI infrastructure push, which includes the Atlas 950 AI SuperPoD designed for trillion-parameter models. Underpinning the platform are Huawei's homegrown technologies, including the GaussDB enterprise-grade database and the DataArts data governance suite. The entire stack runs on Huawei's Da Vinci architecture-based Ascend AI processors, the company's alternative to Nvidia GPUs, which have reportedly been found to contain advanced, stockpiled components from TSMC and SK Hynix. This launch is a piece of a larger strategic vision Huawei calls the "agentic internet era," articulated by senior VP Li Peng. The company is betting on a future where networks primarily connect billions of AI agents, moving beyond human connectivity and requiring a fundamental shift in infrastructure, a strategy they've labeled "5G-A x AI." Alongside performance, Huawei is addressing the operational and energy efficiency of AI workloads. The company is promoting AI-driven green data center solutions to lower power usage effectiveness (PUE) and has also claimed software innovations capable of boosting GPU and NPU utilization from an industry average of 30-40% to as high as 70%.
