NVIDIA AI Aerial
NVIDIA AI Aerial™ is a suite of accelerated computing platforms, software libraries, and tools to build, train, simulate, and deploy AI-native wireless networks. It enables developers and researchers to go from rapid prototyping to commercial development of AI-RAN solutions for 5G and 6G that telcos can deploy.
NVIDIA AI Aerial Software
Explore open source software libraries and tools for building, training, simulating, and deploying AI-native wireless networks. Access to NVIDIA Aerial software, including AODT Client, is available on GitHub. For complete AODT software, access is available through the NVIDIA 6G Developer Program.
NVIDIA Sionna
A GPU-accelerated, differentiable, open source library for 5G and 6G communications research. It features a lightning-fast ray tracer for radio propagation, a link-level simulator, and system-level simulation capabilities.
Read Sionna DocumentationNVIDIA Aerial Framework
A toolchain for generating high-performance, GPU-accelerated 5G/6G pipelines from Python and a modular, real-time runtime for executing the pipelines on NVIDIA Aerial™ RAN Computer platforms.
Read Aerial Framework DocumentationNVIDIA Aerial CUDA-Accelerated RAN
An SDK (Software Development Kit) for building commercial-grade, AI-native, 3GPP, and O-RAN compliant 5G/6G gNB software on NVIDIA-accelerated computing platforms
Go to ACAR GitHub
NVIDIA Aerial Omniverse Digital Twin (AODT)
A platform for creating next-generation network digital twins, enabling physically accurate virtual 5G and 6G wireless networks, from single towers to full cities.
Access AODT Complete (Membership Required)
Read Aerial Omniverse Digital Twin DocumentationHardware Platforms
NVIDIA AI Aerial Research Platforms
These systems enable the building, training, and over-the-air testing of AI-native wireless innovations for 5G and 6G.
Sionna Research Kit
Built on an open source foundation and powered by NVIDIA’s state-of-the-art, GPU-accelerated libraries, the Sionna Research Kit makes rapid prototyping, training, and deployment of cutting-edge 5G and 6G algorithms achievable for everyone—from seasoned professionals to students.
Aerial Testbed (ARC-OTA)
An end-to-end system that includes Aerial CUDA-Accelerated RAN combined with open source software (OAI L2+ and 5G Core) running on NVIDIA GH200 or DGX Spark™, over the air. Used for product development and performance optimization of commercial-grade and software-defined AI-RAN solutions.
NVIDIA AI Aerial Deployment Platforms
The Aerial RAN Computer (ARC) family delivers high-performance, scalable, and accelerated computing platforms for telecom networks, enabling commercial AI-RAN deployments.
Aerial RAN Computer MGX 2U Short
A high-performance, energy-efficient AI-RAN platform in a telco-optimized form factor. ARC MGX 2U Short brings advanced computing, connectivity, and sensing to cell sites, running RAN and AI workloads within a RAN-class power envelope.
Aerial RAN Computer MGX 2U
A modular and high-performance AI-RAN platform in a 2U form factor. Built for AI-centric workloads and designed to scale from distributed RAN (D-RAN) to centralized RAN (C-RAN). Runs AI applications including vision AI, voice AI, and physical AI at the edge.
Aerial RAN Computer MGX 4U
A modular and high-performance AI-RAN platform in a 4U form factor for high-density deployments. Designed for AI-centric workloads and maximizing AI monetization for centralized AI-RAN and AI-only deployments at aggregation sites.
Get Started With NVIDIA AI Aerial
Build and Train
Build and train AI/ML models for RAN, physical (PHY), and media access control (MAC) layers.
Simulate
Run large-scale, photorealistic 5G and 6G wireless network scenarios
Deploy
Implement and validate in live and edge networks at scale.
Learning Library
NVIDIA AI Aerial Ecosystem
Academia and industry leaders are collaborating to advance AI-native wireless networks and drive 6G research.
.svg)
Next Steps
Join the 6G Developer Program
Get access to NVIDIA AI Aerial resources and community.
NVIDIA AI Aerial FAQ
NVIDIA AI Aerial is a suite of accelerated computing platforms, software libraries, and tools for building, training, simulating, and deploying AI-native wireless networks. AI Aerial supports both advanced wireless research and commercial-grade cellular deployment. It enables developers to go from rapid prototyping to commercial development of AI-RAN solutions for 5G and 6G that telcos can deploy.
AI Aerial includes a variety of hardware and software resources for wireless research and commercial deployment. Hardware ranges from small to large accelerated computing platforms, comprising various GPUs, CPUs, and networking components. Software libraries range from open source Sionna for rapid prototyping to Aerial CUDA-Accelerated RAN, Aerial Framework, and Aerial Omniverse Digital Twin for building CUDA-accelerated software-defined RAN.
Developers and researchers can access all NVIDIA Aerial software on GitHub: NVIDIA Sionna, NVIDIA Aerial CUDA-Accelerated RAN, NVIDIA Aerial Framework, and NVIDIA Aerial Omniverse Digital Twin Client. For complete AODT software, access is available through the NVIDIA 6G Developer Program. Documentation, SDKs, and hardware platforms are available for rapid onboarding and experimentation.
Compatible hardware includes Aerial Testbed (ARC-OTA), Aerial RAN Computer MGX 2U Short, Aerial RAN Computer MGX 2U, Aerial RAN Computer MGX 4U, DGX Spark, and the Sionna Research Kit. These are built using various accelerated computing platforms, such as NVIDIA Jetson Orin™, DGX Spark, Grace Hopper 200 for research, and RTX PRO GPUs for commercial deployments.
AI Aerial supports centralized and distributed AI-RAN deployments for public and private 5G and future 6G networks. Aerial CUDA-Accelerated RAN is O-RAN 7.2x compliant and supports O-RAN Distributed Unit (O-DU) functionality.
For specific questions about NVIDIA Aerial CUDA-Accelerated RAN, please ask in the GitHub issues in the repo. For specific questions about NVIDIA Aerial Frameworks, please ask in the GitHub issues in the repo. For all other questions, please ask in the Aerial Developer Forum.