Huawei’s AI Data Platform was unveiled at the Huawei AI DC Innovation Forum during MWC Barcelona 2026 on March 5, 2026, targeting key barriers that prevent enterprise artificial intelligence agents from moving beyond proof-of-concept stages.

Xie Liming, President of the Flash Storage Domain at Huawei’s Data Storage Product Line, presented the platform in Barcelona. The platform integrates three core components: a knowledge base, a key-value (KV) cache, and a memory bank, coordinated through a Unified Cognitive Management (UCM) layer.

The Problem: Why AI Agents Stall at the Demo Stage

Enterprises today hold large volumes of data but struggle to deploy AI agents at scale. Huawei identified several persistent obstacles, including delayed knowledge acquisition, low retrieval accuracy, and inefficient inference in long-sequence and multi-role interaction scenarios. Moreover, the absence of task memory and accumulated experience keeps most AI agents confined to demonstrations rather than operational use.

AI Data Platform: Three Integrated Capabilities

The platform addresses these gaps through three distinct technical functions. Each function targets a specific failure point in enterprise AI agent deployment.

  • Knowledge Generation and Retrieval: The knowledge base monitors changes in source data continuously and converts raw data into knowledge in near real time. It processes multimodal data through lossless multimodal parsing and token-level encoding, achieving retrieval accuracy above 95%.
  • KV Cache for Inference Acceleration: Intelligent tiering and management of the KV cache reduce redundant computation during inference. This lowers latency, improves inference throughput, and supports performance in long-sequence and complex agent reasoning scenarios.
  • Memory Extraction and Recall: The memory bank accumulates working memory and experience memory during AI agent interactions. It supports memory recall and collaborative learning across multiple agents, enabling models to improve reasoning accuracy and efficiency over time.

Context: Enterprise AI Adoption Challenges

The launch reflects a broader industry challenge. As AI adoption accelerates across sectors, the gap between laboratory performance and real-world enterprise deployment remains significant. Huawei’s platform targets this gap directly by addressing data infrastructure limitations rather than model architecture alone.

The computing demands of long-context inference and multi-agent coordination have placed increasing pressure on storage and memory systems. Huawei’s approach combines storage-layer intelligence with memory management to reduce that pressure at the infrastructure level.

Future Outlook

Huawei stated it plans to increase investment in AI data infrastructure and will continue working with customers and partners globally to expand AI adoption across additional industries. The company said these efforts aim to unlock the full potential of enterprise data assets.

“The platform enables AI agents in enterprises to move beyond the demo stage and become actual operational tools.”

Xie Liming, President of Flash Storage Domain, Huawei Data Storage Product Line