MWC Barcelona Kicks Off With AI Focus

The discourse at MWC has decisively shifted from AI experimentation to full-scale, production-grade infrastructure. The focus is now on AI-RAN (Radio Access Network), where AI and machine learning are embedded directly into the network fabric to automate planning, optimize operations, and manage the exponential growth in machine-driven data traffic. This move to "AI-native" systems is seen as a foundational step toward 6G. Agentic AI is the dominant theme, with a clear push towards Level 4 autonomous networks that can self-heal and run with zero-touch. Deutsche Telekom, in partnership with Google Cloud, is showcasing MINDR (Multi-Agentic Intelligent Network Diagnostics & Remediation), a multi-agent system built on Gemini models. An early version focused on the RAN has already reduced the time to manage major network events from hours to minutes, a greater than 95% improvement. On the hardware front, GPU-accelerated infrastructure is central. ZTE unveiled its SuperPOD, a high-density compute platform supporting up to 128 GPUs in a single rack with an innovative cable-free design. Meanwhile, the AI-RAN ecosystem is heavily leveraging NVIDIA's platforms, with partners like Nokia and Samsung demonstrating vRAN software running concurrently with AI workloads on NVIDIA Grace Hopper Superchips. For developers and MLOps teams, the tooling is becoming more accessible. ZTE's Co-Sight AI Agent Studio offers a low-code, visual platform for building industrial-grade agents, featuring modular components and a repository for models and tools. This platform supports creating Retrieval-Augmented Generation (RAG) knowledge bases and uses a Directed Acyclic Graph (DAG) mechanism to intelligently parallelize tasks for more efficient execution. The underlying cloud infrastructure is also evolving to support these distributed AI workloads. Huawei Cloud's Foundation (HCF) is being positioned for hybrid environments, and its Cloud Container Engine (CCE) is a Kubernetes-certified service. Google Cloud is contributing to the open-source ecosystem by releasing its telco data pipeline and data models on GitHub, aiming to standardize automation based on Kubernetes and the Nephio project. This integration of AI into core network services is forcing a re-evaluation of commercial models. The industry is moving away from simple subscription or per-seat licenses. The emerging consensus points towards hybrid, usage-based, and outcome-based pricing, where costs are directly tied to consumption metrics like API calls, data processed, or specific business results achieved.