Huawei Launches AI Data Platform to Bridge AI Models and Business Value

Barcelona, Spain — At Huawei’s Product & Solutions Launch event during MWC Barcelona 2026, Yuan Yuan, President of Huawei Data Storage Product Line, officially introduced the Huawei AI Data Platform. The platform integrates technologies including knowledge generation and retrieval, KV cache to accelerate inference, extraction and recall of memory, and the Unified Cache Manager (UCM). The solution aims to bridge the gap between AI model capabilities and real business value.

Although AI development is progressing rapidly, many AI models are not yet fully embedded in a company’s core services. This is because development focus has traditionally been on training rather than inference. Yet, inference is key to widespread application of AI models. In addition, data processing improvements are required to address challenges such as AI hallucinations, slow responses, and limited throughput during inference.

To address these challenges, Huawei Data Storage launched the AI Data Platform “3+1”, which comprises:

• Knowledge generation and retrieval with highly accurate multimodal knowledge

This platform converts various multimodal data sources into a high-accuracy knowledge base through multimodal parsing without data loss, and token-level programming, achieving search accuracy of over 95%.

• KV cache that accelerates inference using historical memory data

This technology employs intelligent management and clustering of KV caches to broaden the context window and reduce repeated computations. As a result, Time to First Token (TTFT) is reduced by up to 90%, making AI responses significantly faster.

• Memory extraction and recall with context memory management

The platform accurately extracts historical data and experiences, storing them as memory that can be recalled. This mechanism makes AI models smarter with continued use.

• UCM for managing memory data across the usage cycle

UCM manages memory data across three cache layers to intelligently orchestrate the knowledge base, KV cache, and memory bank.

Huawei AI Data Platform offers two deployment modes: appliance mode for new implementations and an independent mode for integrating with existing systems. In appliance mode, the platform uses the OceanStor A800 system to deliver high performance and scalable flexibility. Meanwhile, the independent mode uses an architecture of AI data engine nodes plus OceanStor Dorado storage devices. The data nodes can be added to scale existing systems, protect prior investments, and support a smoother AI transformation for enterprises.

“We will continue to deepen technological innovation. With the AI Data Platform as a bridge, we will convert AI model capabilities into real business value,” Yuan Yuan said.

Press contact:

Zoe Yin, yinzijun@huawei.com

SOURCE Huawei