MWC Barcelona 2025

AI-RAN Alliance at Mobile World Congress 2025

At MWC Barcelona 2025, the AI-RAN Alliance showcased cutting-edge demonstrations that highlighted AI’s transformative potential across wireless networks. This page serves as a central hub for all featured demos—now available as on-demand recordings—offering a comprehensive look at how AI-native RAN is shaping the future of connectivity. Explore real-world applications of AI that optimize RAN performance, boost energy efficiency, and unlock new levels of automation.

AI-RAN at MWC Barcelona 2025: Demo Highlights & Innovation in Action

Relive the standout moments from the AI-RAN Alliance’s presence at Mobile World Congress Barcelona 2025. This video captures key demos and innovations that show how AI is redefining the Radio Access Network—improving performance, boosting energy efficiency, and enabling smarter automation.

Demo 1: “Learned Air Interface with Online Learning”

Description: This demo shows how AI-driven air interface design can enhance wireless performance, improve spectral efficiency, and integrate seamlessly with existing 5G networks, making it a critical capability for future AI-RAN deployments.

  • Participants: DeepSig and NVIDIA
  • Category: AI-for-RAN

Demo 2: “Realization of UL Ch Interpolation in Actual RAN”

Description: The demo showcases uplink channel interpolation in a real-world Radio Access Network (RAN), focusing on estimating and optimizing wireless communication for better performance and efficiency.

  • Participants: Fujitsu, NVIDIA, and SoftBank
  • Category: AI-for-RAN

Demo 3: “AI-based PUSCH Channel Estimation”

Description: Using an AI-powered uplink (UL) channel estimation (CE), this demo showcases how the uplink throughput can be improved by over 30% in 5G and future 6G networks. We evaluate this novel solution in a real-time commercial testbed with a Samsung developed AI CE algorithm running on a NVIDIA Grace-Hopper GPU server(vDU), Keysight Propsim Channel Emulator and Core Emulator and a Samsung commercial UE. Samsung’s AI algorithm leverages the multi-dimensionality of the received DMRS to enhance channel estimation and hence PUSCH performance, providing a better user experience and a coverage extension for operators.

  • Participants: Keysight, NVIDIA and Samsung
  • Category: AI-for-RAN

Demo 4: “AI/ML Optimized Higher-Order Modulations with a Neuromorphic Receiver”

Description: Samsung, NVIDIA and VIAVI will highlight the transformative potential of AI in RAN. Samsung will showcase AI/ML-based higher-order modulations to deliver improved performance over square QAMs. VIAVI will showcase energy-efficient neuromorphic receivers that replace several signal-processing blocks—channel estimation, equalization, and symbol de-mapping—with a single neural network, optimized for AI-based modulations. NVIDIA will feature its high-performance GPUs for training and validation.

  • Participants: NVIDIA, Samsung, and VIAVI
  • Category: AI-for-RAN

Demo 5: “AI-based 5G Beamforming for Mobility-Aware Interference Mitigation and Power Saving”

Description: This demo is for AI-powered beamforming and energy-efficient RAN control in 5G networks. By leveraging realistic mobility data and hierarchical RL, it showcases how AI can enhance interference management, optimize power use, and improve network efficiency in real-time mobility scenarios.

  • Participants: Singapore University of Technology (SUTD), VIAVI, and Yonsei University
  • Category: AI-for-RAN

Demo 6: “AI-based Spectrum Sensing in the RAN”

Description: This demo highlights the potential of AI-enhanced RAN sensing for dynamic spectrum sharing, showing how AI can enable smarter interference avoidance, improved spectral efficiency, and adaptive radio resource management in real-world 5G deployments.

  • Participant: Northeastern University Open6G
  • Category: AI-for-RAN

Demo 7: “AI-RAN Orchestration”

Description: This demo showcases AI-driven RAN orchestration, highlighting how AI and RAN can coexist on a shared infrastructure while maintaining high performance, efficiency, and quality of service.

  • Participants: Keysight and Northeastern University Open6G
  • Category: AI-and-RAN

Demo 8: AI-Driven Spectrum Sensing for Dynamic & Privacy-preserving AI Model Partitioning over 5G Network

Description: This demo highlights adaptive AI-driven ML model partitioning for privacy-focused image processing over 5G networks, addressing the inefficiencies of fixed model partitioning in dynamic wireless environments.

  • Participants: LITEON, Keysight, NeuroRAN and Singapore University of Technology and Design (SUTD)
  • Category: AI-on-RAN

Demo 9: “Integrated Sensing and Communications (ISAC)”

Description: This demo highlights the use of integrated sensing and communications (ISAC) over existing 5G networks. By repurposing a commercial 5G waveform as a radar signal, it can detect and track unconnected objects—such as pedestrians—without relying on cameras or RF tags. Potential commercial applications include occupancy sensing, drone detection, and perimeter security.

  • Participants: Northeastern University and Tiami Networks
  • Category: AI-on-RAN

Demo 10: “AI on RAN Object Detection”

  • Description: This demo showcases a powerful, cost-effective solution for deploying AI workloads on Private 5G networks, making them more accessible for industries requiring real-time processing, scalability, and low latency.
  • Participants: Arm, Effnet AB*, Phluido*, and Tannera*
  • Category: AI-on-RAN