NVIDIA has expanded its Aerial Research portfolio with new tools aimed at accelerating AI-native wireless networks and 6G research, as announced on March 18, 2025. The tools empower developers and telecom leaders to innovate in AI-RAN technology.
The telecom industry is increasingly integrating AI to ensure seamless connectivity in areas with poor signal strength while enhancing sustainability and spectral efficiency. AI-RAN advancements are paving the way for AI-native wireless networks designed to cater to billions of connected devices, sensors, robots, cameras, and autonomous vehicles.
New offerings in the NVIDIA Aerial Research portfolio
The newly launched tools include the Aerial Omniverse Digital Twin (AODT) on NVIDIA DGX Cloud, the Aerial Commercial Test Bed (ARC-OTA) on NVIDIA MGX, the NVIDIA Sionna 1.0 open-source library, and the Sionna Research Kit on NVIDIA Jetson. These offerings are tailored to support AI-RAN and 6G development efforts.
The Aerial Omniverse Digital Twin provides a simulation platform for testing algorithms in detailed digital replicas of wireless systems, accessible via multiple environments including on-premises and cloud options. The Aerial Commercial Test Bed allows for real-time deployment and testing of AI models over the air, incorporating commercial-grade RAN software and open-source elements to form a comprehensive AI-RAN testing ecosystem.
Sionna 1.0 has become the most downloaded GPU-accelerated open-source library for communication systems, now featuring a ray tracer for radio propagation and advanced simulation capabilities. The Sionna Research Kit supports researchers in prototyping AI-RAN algorithms by enabling quick connections to 5G equipment.
The NVIDIA Aerial Research portfolio is further bolstered by the NVIDIA 6G Developer Program, which engages over 2,000 members from leading tech companies, academia, and telecom operators dedicated to advancing AI-RAN and 6G research.
Llama Nemotron: Nvidia’s answer to the AI reasoning boom
Industry impact and collaborations
Testing and simulation are crucial for the development of AI-native wireless networks. Companies including Amdocs, Ansys, and Keysight are improving their simulation solutions through NVIDIA’s AODT. The AI-RAN Alliance highlighted that nine out of ten demonstrations presented at Mobile World Congress utilized the NVIDIA Aerial Research tools, yielding significant advancements.
In particular, SoftBank and Fujitsu achieved a 50% throughput increase in suboptimal radio conditions by applying AI-based uplink channel interpolation. DeepSig developed OmniPHY, which leverages neural networks to eliminate traditional overhead, facilitating up to 70% throughput gains. This AI-native air interface integrates machine learning into critical components to optimize spectral efficiency and reduce power consumption.
“AI-native signal processing is transforming wireless networks, delivering real-world results,” stated Jim Shea, cofounder and CEO of DeepSig, emphasizing the significant impact of integrating deep learning into wireless network design.
The ecosystem of NVIDIA CUDA-X libraries also supports developers in creating high-performance applications to further explore AI in telecommunications.
Featured image credit: Nvidia
