Barcelona, Spain – Major telecommunications companies are moving AI-powered radio access networks (AI-RAN) from laboratory testing to real-world deployments, with significant milestones announced this week at Mobile World Congress (MWC). Nokia and NVIDIA are at the center of these advancements, forging modern collaborations with operators across Europe, Asia, and North America.
T-Mobile U.S. Has demonstrated concurrent AI and RAN processing on an NVIDIA AI-RAN platform utilizing Nokia’s CUDA-accelerated RAN software. In field tests, Nokia’s AirScale massive MIMO radio operating in the 3.7GHz band supported commercial devices running applications including video streaming, generative AI, and AI-powered video captioning alongside standard 5G services.
SoftBank achieved an industry first with its AITRAS live field trial, successfully implementing a 16-layer massive MIMO system using fully software-defined 5G powered by NVIDIA’s AI-RAN platform. This represents a key technical step toward the commercialization of AI-RAN technology.
Indosat Ooredoo Hutchison (IOH) has progressed from proof-of-concept to pre-commercial field validation, implementing software-defined 5G with Nokia’s vRAN software on NVIDIA AI-RAN platforms. At MWC, IOH showcased Southeast Asia’s first AI-powered 5G call, demonstrating secure, real-time cross-border connectivity and remote control of a robotic dog over a live 5G network.
Benchmarking results from SynaXG indicate that AI-RAN running on NVIDIA platforms delivers high-speed, carrier-grade performance across multiple 5G spectrum bands. SynaXG likewise demonstrated a fully software-defined AI-RAN implementation using NVIDIA AI Aerial, achieving a throughput of 36 Gbps with under 10 milliseconds latency on a single NVIDIA GH200 server, and activating 20 component carriers with a centralized and distributed unit on one platform.
The number of AI-RAN innovations showcased at MWC this year has tripled compared to last year, with 26 out of 33 demos from the AI-RAN Alliance built using NVIDIA AI Aerial and a software-defined architecture. These demonstrations highlight the potential of AI to boost 5G performance and unlock new edge AI applications.
Several companies are demonstrating specific AI-RAN applications. DeepSig is working to improve signal formats using AI, potentially increasing throughput and efficiency. SUTD, NVIDIA, and partners are exploring how to distribute AI processing across devices, edge computing, and the cloud for autonomous systems. ZTouch Networks and partners have developed an AI-RAN orchestration blueprint to share GPUs efficiently between AI and RAN workloads. Northeastern University and SoftBank are demonstrating an AI switching solution that dynamically selects between AI and traditional algorithms for channel estimation.
Capgemini, through Project ULTIMO, is demonstrating how AI-RAN can support large-scale autonomous mobility services, enabling autonomous shuttles to process sensor data locally and leverage AI applications on NVIDIA AI-RAN servers for scene understanding and safety detection.
An expanding ecosystem of partners is supporting NVIDIA-powered AI-RAN platforms. Quanta Cloud Technology (QCT) is releasing commercial off-the-shelf AI-RAN products, while Supermicro is extending support across the NVIDIA AI-RAN portfolio. WNC has introduced a new AI-optimized radio unit, and LITEON has integrated its radio units with NVIDIA AI Aerial.
NVIDIA has open sourced NVIDIA Aerial CUDA-accelerated RAN libraries and has joined the OCUDU (Open CU DU) Ecosystem Foundation, contributing to open source RAN software development. According to NVIDIA’s latest State of AI in Telecom report, 77% of respondents anticipate a faster deployment of AI-native wireless network architecture than previously expected.
