NVIDIA has launched AI Aerial, an advanced AI-driven platform designed to meet the evolving demands of next-generation telecommunications networks. This platform offers telecom providers a full suite of capabilities, combining cutting-edge software and hardware for developing, training, and deploying AI radio access network (AI-RAN) technology. The primary aim of AI Aerial is to optimize wireless networks, improving efficiency and performance while reducing the total cost of ownership. It also opens new revenue opportunities for telecom operators, particularly as AI expands into areas such as autonomous vehicles, generative AI, smart factories, and robotics.
At the core of NVIDIA AI Aerial is its support for dynamic, AI-based workload allocation across 5G and 6G networks, increasing capacity utilization by 2-3 times. The platform integrates CUDA-Accelerated RAN, which provides access to high-performance software libraries designed to enable the development and deployment of virtualized RAN workloads on NVIDIA’s accelerated compute platforms. This ensures the efficient handling of complex networking functions, especially for AI-driven services. By utilizing AI Radio Frameworks—based on PyTorch and TensorFlow—telecom operators can develop, train, and deploy AI models that improve spectral efficiency and introduce new capabilities to radio signal processing for 5G and future 6G systems. This suite also includes NVIDIA Sionna, a tool specifically designed for simulating and training neural network-based 5G/6G algorithms, improving network performance through AI-enhanced signal processing.
NVIDIA Omniverse Digital Twin (AODT) is another standout feature within AI Aerial, providing telecom operators with a digital twin simulation environment for testing network performance in real-world scenarios. By replicating network systems, from a single base station to city-wide deployments, operators can optimize network coverage and ensure reliability across various terrains and environments. The Aerial-CUDA Accelerated RAN capabilities in conjunction with AODT allow for simulation and testing of user equipment (UE) and other network components, facilitating a seamless transition from development to real-world deployment. This level of simulation provides valuable insights into how networks perform in specific environments, supporting more accurate planning and optimization of telecom infrastructures.
One of the key benefits of AI Aerial is its ability to integrate AI directly into the core of network operations, allowing telecom providers to host not only traditional RAN traffic but also AI-driven applications, including generative AI and robotic teleoperations. AI-RAN is designed to support a wide range of next-generation use cases, including robotic surgery, 3D collaboration, spatial computing, and autonomous vehicle management. As these applications demand higher network performance, AI Aerial’s ability to provide software-defined, scalable RAN solutions becomes crucial for maintaining low-latency, high-speed connectivity.
As part of its broader efforts to commercialize AI-RAN technology, NVIDIA has partnered with T-Mobile, Ericsson, and Nokia to establish the AI-RAN Innovation Center, which will focus on driving advancements in AI-RAN capabilities. This center aims to accelerate the adoption of AI in wireless networks by combining NVIDIA’s AI Aerial platform with the telecom expertise of its partners. Through this collaboration, the AI-RAN Innovation Center will foster the development of AI-powered telecom infrastructure that can meet the demands of the AI era.
In addition to telecom operators, the NVIDIA AI Aerial ecosystem includes key partners such as Softbank, Fujitsu, Ansys, Keysight, and a range of academic institutions like ETH-Zurich and Northeastern University, all collaborating on 6G research and AI-driven advancements. The platform’s ability to support diverse deployment scenarios—from O-RAN to private 5G and virtual RAN (vRAN)—enables flexibility in network scaling and resource optimization. Moreover, AI Aerial’s software-defined architecture ensures that telecom providers can transition from 5G to 6G with a simple software upgrade, future-proofing networks and reducing the need for costly infrastructure overhauls.
• NVIDIA AI Aerial integrates AI into 5G and 6G RAN optimization, enhancing spectral efficiency and performance.
• The platform includes CUDA-Accelerated RAN and AI Radio Frameworks for developing AI-driven networking solutions.
• Omniverse Digital Twin provides a detailed simulation environment for network deployment and testing.
• AI Aerial supports dynamic workload allocation, improving energy efficiency and resource utilization by 2-3 times.
• Collaboration with T-Mobile, Ericsson, and Nokia via the AI-RAN Innovation Center to accelerate AI-RAN commercialization.
“AI-RAN is set to revolutionize the telecom industry, and the opening of the AI-RAN Innovation Center will help to take us on this journey by driving industry collaboration,” said Tommi Uitto, president of Mobile Networks at Nokia. “By bringing together leading companies in the telecom and AI industries, we can unlock the full potential of AI in our networks, improving performance, reducing costs and creating new opportunities for our customers. We are confident that AI-RAN will be a key driver of innovation in the future, and we are excited to be part of this revolution together with NVIDIA.”
“Ericsson has invested in our AI-RAN solution, allowing communications service providers to deploy portable RAN software running across multiple platforms. We are now evaluating the performance and cost of NVIDIA accelerated computing in this context,” said Fredrik Jejdling, EVP and head of Business Area Networks at Ericsson
