SynaXG, a Singapore-based technology company, has unveiled a significant advancement in AI-native radio access networks (AI-RAN) at the Mobile World Congress 2026 in Barcelona. The company demonstrated the concurrent operation of 5G FR1 and FR2 with AI workloads on the NVIDIA AI Aerial platform, showcasing a fully software-defined AI-RAN architecture that delivers carrier-grade performance without the need for separate hardware.
The demonstration highlighted several key achievements. SynaXG achieved over 36Gbps throughput across 20 x 100 MHz 5G cells with sub-10 millisecond latency on a single NVIDIA GH200, matching top commercial 5G deployments. This marks the first-ever carrier-grade FR2 virtualised RAN running alongside FR1 and AI workloads on a shared GPU, crucial for dense urban and enterprise 5G applications.
SynaXG’s system operated continuously under sustained load, proving its stability and readiness for commercial deployment. The real-time orchestration software dynamically balanced GPU resources between RAN and AI workloads, ensuring uninterrupted service and optimising infrastructure use. “With the recent industry-leading breakthroughs, SynaXG has demonstrated that AI-RAN can deliver carrier-grade FR1 and FR2 performance with continuous 24×7 operation on shared NVIDIA AI infrastructure,” said Xin Huang, CEO of SynaXG.
Looking forward, SynaXG plans to extend its AI-RAN platform to additional NVIDIA platforms, including NVIDIA DGX Spark, to further enhance its software-defined architecture. This innovation promises to accelerate deployment and improve network optimisation, energy efficiency, and operational simplicity in AI-native networks.
