[Barcelona, Spain, March 2, 2026] At the Huawei Product & Solution Launch 2026 at MWC Barcelona 2026, Yuan Yuan, President of Huawei Data Storage Product Line, officially announced Huawei’s AI data platform. The platform integrates three key technologies: knowledge generation and retrieval, memory extraction and invocation, KV cache for inference acceleration, and a unified cache manager (UCM) for inference memory. Bridging the gap between model and business value.
Mr. Yuan Yuan, President of Huawei Data Storage Product Line, giving the keynote speech
Despite rapid advances in AI, most AI models are not fully integrated into carriers’ core services. This is primarily due to the focus on training rather than inference, which is the key to model adoption. Improvements in data processing are needed to address challenges such as AI hallucinations, slow response times, and throughput constraints in inference.
Huawei Data Storage has officially launched a “3+1” AI data platform that tackles these challenges through three breakthrough technologies and software:
- Knowledge generation and retrieval with high-precision multimodal knowledge for more accurate retrieval
For smart search applications, this technology transforms multimodal resources (text, images, videos, etc.) into high-precision knowledge through multimodal lossless parsing and token-level encoding, ensuring search accuracy of over 95%.
- Memory extraction and recall using contextual memory management makes your models smarter as you use them.
In business data insight scenarios, this technology allows you to define problems and understand user intent through personalized memory when faced with a task. Then use episodic memory (or experience-based memory) to break down the task. This memory mechanism allows your model to get smarter as you use it.
- Faster and more efficient inference using historical memory data to accelerate inference with KV cache
In AI customer service scenarios, the technology applies intelligent tiering and management of KV caches to significantly expand the context window and reduce repetitive computation. This reduced time to first token (TTFT) by 90% and significantly improved AI response speed.
- UCM for managing and scheduling memory data throughout its lifecycle
The software manages memory data at three cache levels and intelligently schedules knowledge bases, memory banks, and KV caches.
Huawei AI Data Platform offers appliance mode for greenfield deployments and independent mode for evolving deployments. Appliance mode uses the OceanStor A800 system, which integrates three key technologies, computing power, and applications. This ensures ultimate performance and flexible scalability. Independent mode uses an architecture of AI Data Engine nodes + OceanStor Dorado storage. You can add Data Engine nodes to upgrade your existing systems, protect your previous investments, and enable a smooth AI transformation for your business.
“Huawei will continue to deepen its technological innovation. With the AI data platform acting as a bridge, we will turn model capabilities into real business value and work with all stakeholders to embrace the future of intelligent computing,” Yuan Yuan said.
MWC Barcelona 2026 will be held from March 2nd to March 5th in Barcelona, Spain. During the event, Huawei will be showcasing its latest products and solutions for 1 hour and 50 minutes at the Fira Gran Via Hall 1 stand.
Now, the era of agent networks is rapidly approaching and large-scale commercial deployment of 5G-A is accelerating. Huawei is actively working with carriers and partners around the world to realize the full potential of 5G-A and pave the way for the evolution to 6G. We also build AI-centric network solutions that enable intelligent services, networks, and network elements (NEs), accelerate large-scale deployments of Level 4 Autonomous Networks (AN L4), and upgrade core businesses with AI. Together with other industry players, we will build a cutting-edge, value-driven network and AI computing backbone for a fully intelligent future.
For more information, please visit https://carrier.huawei.com/en/minisite/events/mwc2026/.
