NVIDIA and its partners are demonstrating that software-defined AI-RAN is the foundation for the next generation of wireless networks, moving from lab experiments to real-world deployments ahead of Mobile World Congress in Barcelona. The advancements highlight a shift toward AI-native, open, and secure 6G-ready systems powered by accelerated computing. Leading telecom operators including T-Mobile U.S., SoftBank, and Indosat Ooredoo Hutchison (IOH) have achieved key milestones using NVIDIA AI-RAN platforms. T-Mobile conducted live over-the-air tests with Nokia’s CUDA-accelerated RAN software, successfully running AI and RAN workloads simultaneously on the 3.7GHz band, supporting applications like video streaming, generative AI, and AI-powered video captioning on commercial devices. SoftBank completed a groundbreaking field trial using a fully software-defined 5G network on NVIDIA’s AI-RAN platform, achieving the industry’s first 16-layer massive MIMO configuration — a major step toward commercializing AI-RAN. IOH implemented software-defined 5G using Nokia’s vRAN software on NVIDIA platforms, advancing from proof of concept to pre-commercial validation. At MWC, they showcased Southeast Asia’s first AI-powered 5G call, enabling real-time, secure cross-border connectivity and remote control of a robotic dog — a demonstration of intelligent, low-latency network capabilities. SynaXG delivered a world-first by deploying AI-RAN on millimeter wave (FR2) spectrum using NVIDIA AI Aerial. Their setup ran 4G, 5G (sub-6GHz and FR2), and agentic AI workloads on a single NVIDIA GH200 server, activating 20 component carriers with both centralized and distributed units. The system achieved 36 Gbps throughput and under 10 milliseconds latency — proof of carrier-grade performance for AI-RAN in demanding environments. This year’s MWC features a threefold increase in AI-RAN innovations compared to last year, with 26 of 33 AI-RAN Alliance demos built on NVIDIA AI Aerial and software-defined architectures. Notable demonstrations include: DeepSig’s AI-native air interface, where neural networks jointly learn signal encoding and decoding, reducing pilot overhead and improving spectral and energy efficiency by up to 2x. SUTD, NVIDIA, and partners showing split-inference for autonomous systems, distributing AI processing across devices, edge, and cloud to meet strict latency and privacy requirements. zTouch Networks presenting a GPU orchestration blueprint using NVIDIA Multi-Instance GPU technology, enabling safe, real-time sharing of GPU resources between AI and RAN workloads — a key enabler for multi-tenant AI-RAN networks. Northeastern University and SoftBank demonstrated a real-time AI switching solution that dynamically selects between AI and classical algorithms for channel estimation, optimizing performance and stability under varying conditions. “AI-RAN is emerging as a unifying architecture for future radio networks,” said Alex Choi, chair of the AI-RAN Alliance. “By aligning operators, vendors, and researchers around software-defined, GPU-accelerated systems, we’re accelerating innovation and laying the groundwork for AI-native 6G.” AI-RAN is critical for autonomous systems like robots and self-driving vehicles, which rely on real-time, intelligent connectivity. Capgemini’s Project ULTIMO, funded by Horizon Europe, is using NVIDIA Jetson Orin modules in autonomous shuttles to process sensor data locally, while critical video and telemetry streams are sent over 5G to AI-RAN servers for scene understanding, safety detection, and accessibility analysis — with mission-critical traffic prioritized. NVIDIA’s AI-RAN ecosystem is growing rapidly, supported by platforms like the NVIDIA Aerial RAN Computer, which combines the Grace CPU with GPUs for high-performance, energy-efficient computing. NVIDIA has open-sourced its AI Aerial CUDA-accelerated RAN libraries and joined the OCUDU (Open CU DU) Ecosystem Foundation to advance open-source RAN development. A recent NVIDIA State of AI in Telecom report reveals that 77% of telecom leaders expect faster deployment of AI-native RAN and 6G systems, signaling a pivotal shift ahead of the traditional 6G timeline. These developments mark a decisive move toward secure, open, and AI-native wireless infrastructure.