5G and 6G Solutions
Deploy 5G and 6G telecom networks that can handle voice, data, video, AI, and generative AI workloads on one common infrastructure.
Read AI-RAN Blog | For Developers
Overview
In addition to voice, data, and video, networks now need to support AI traffic generated by smartphones, cameras, robots, drones, smart glasses, and more. This shift creates a transformative opportunity for communication service providers (CSPs), who can use their existing networks to deliver AI inferencing at scale with low latency and guaranteed quality of service (QoS). NVIDIA AI Aerial makes it possible, offering the software-defined, accelerated platform that can power RAN and AI from the same infrastructure, so CSPs can support AI today while future-proofing for 6G.
Telecom industry leaders are pioneering a unified solution called AI-RAN, harnessing the power of NVIDIA AI Aerial, to transform cellular networks for the AI era.
Supporting a growing ecosystem for AI-RAN, the new platform will transform radio access network technology into AI-driven computing infrastructure.
Benefits
NVIDIA AI Aerial platform provides accelerated compute for new AI applications and services. CSPs can monetize this today with AI factories, and host their own generative AI applications or leverage NVIDIA AI Enterprise to build new AI offerings.
NVIDIA AI Aerial platform supports dynamic allocation of 5G and AI workloads based on demand within the same GPU. This increases capacity utilization by 2–3X, improves energy efficiency and space utilization, and reduces operational overhead of siloed infrastructure.
NVIDIA Aerial AI Radio Frameworks provide advanced layer 1 (L1) and layer 2 (L2) AI algorithms to improve spectral efficiency by up to 20 percent, which means CSPs can take on more subscribers and improve QoS.
With NVIDIA AI Aerial platform, CSPs can move to 6G with a software upgrade, thanks to a fully software-defined architecture. It’s already being used as a key part of 6G research and development.
The NVIDIA AI Aerial platform supports different market segments, including O-RAN, private 5G, and virtual RAN (vRAN); different workloads; and different deployment scenarios, from radio access, metro edge, and core to central cloud.
Key Components
The NVIDIA AI Aerial platform is a comprehensive suite of hardware and software components, underpinned by a fully software-defined architecture that enables easy scale-up, scale-out, and multi-tenancy to deliver AI and RAN together.
Aerial CUDA®-Accelerated RAN offers highly programmable 5G RAN acceleration libraries for L1 and L2+ layers.
With GPU-accelerated processing, complex computations run faster than on non-GPU solutions.
It’s scalable and deployable on the same infrastructure as generative AI applications.
Aerial AI Radio Frameworks provide a package of AI algorithms to enable AI training and inference in the RAN.
They offer seamless integration with popular frameworks like PyTorch and TensorFlow, leveraging NVIDIA GPUs for generating and capturing data and training AI and machine learning models for RAN at scale.
NVIDIA Aerial Omniverse™ Digital Twin enables physically accurate simulations of wireless systems, from a single tower to city scale.
It enables developers to simulate and build base-station algorithms based on site-specific data and train models in real time to improve transmission efficiency.
NVIDIA Spectrum™ switches can be dynamically programmed to route traffic to any virtual distributed unit (vDU) deployed on any server.
Spectrum combines all functionality needed to run 5G and AI on the same infrastructure with adaptive routing.
It supports advanced timing protocols like Precision Time Protocol, Synchronous Ethernet, and packets per second, as well as dynamic radio unit (RU) and DU mapping.
AI-RAN Orchestrator dynamically allocates compute resources within or across GPUs to optimize resource usage and lower operational costs.
AI-powered automation streamlines network operations, reduces human error, and enhances efficiency.
NVIDIA Cloud Functions (NVCF) enable serverless APIs for AI workload orchestration, which helps network operators create new revenue sources.
Use Cases
Explore how a multi-purpose, accelerated AI-RAN infrastructure supports current and future industry use cases, driving business outcomes for telcos.
The NVIDIA AI Aerial platform enables software-defined RAN by leveraging GPU acceleration and AI-driven analytics to optimize network performance. With this foundation, CSPs can use the NVIDIA AI Aerial platform to deploy any virtual RAN variant, including 5G vRAN, Open RAN (O-RAN), and private 5G.
Independent software vendors can also take the NVIDIA AI Aerial CUDA-accelerated vRAN implementation and customize it with their own enhancements to create new vRAN products for different market segments and deployment scenarios.
From improving customer experiences and optimizing complex network operations with generative AI to extracting business insights with data science, CSPs are unlocking new ways to positively impact their bottom line. The NVIDIA AI Aerial platform lets generative AI applications run on the same AI-RAN infrastructure and support CSPs’ internal use cases.
With NVIDIA AI Enterprise and NVIDIA NIM™, a suite of prepackaged AI modules and libraries, CSPs can also build their own AI and generative AI offerings and deliver them to their customers over a common AI-RAN infrastructure.
CSPs are perfectly positioned to expand their data centers with accelerated computing infrastructure that can support sovereign AI models and manufacture intelligence. This new class of data centers, known as AI factories, empower local governments, enterprises, and startups to build, develop, and deploy generative AI applications with enhanced security, performance, and fine-tuning for local language and cultural standards.
With the NVIDIA AI Aerial platform, CSPs can turn their small sites into mini AI factories and monetize them with GPU-as-a-service offerings, particularly for running inference at the edge.
The NVIDIA AI Aerial platform can be used to accelerate any containerized network function, including vRAN DU, centralized unit (CU), user plane function (UPF), radio intelligent controller (RIC), virtual router (vRouter), network security, and more. This approach maximizes resource utilization and energy efficiency and reduces total cost of ownership (TCO) compared to running siloed infrastructure for each network function.
The extensible architecture also allows ecosystem partners or CSPs to bring their own containerized network functions, leveraging accelerated compute and software stack from the AI Aerial platform.
The NVIDIA AI Aerial platform enables 6G research at all layers in the RAN stack. The platform includes tools that support training and inference, assembling datasets from real-world and synthetic data, and simulation spanning from link level to city scale. New algorithms and AI and machine learning in L1 and L2 can be rapidly developed, simulated, and verified in the real world using a real-time, over-the-air testbed.
Ecosystem
Leaders from across the telecom industry are collaborating on cellular networks for the AI era.
Resources
Browse the latest news, blogs and demos.
Harnessing the power of NVIDIA AI Aerial, telecom industry leaders are pioneering a unified solution called AI-RAN to transform cellular networks for the AI era.
NVIDIA is introducing Aerial RAN Computer-1, the world’s first AI-RAN deployment platform, that can serve AI and RAN workloads concurrently, on a common accelerated infrastructure.
Request a consultation with NVIDIA telecom experts to explore how the NVIDIA AI Aerial platform can transform your network.
Ask questions or request more information. Our experts are here to help.
Register to get the latest AI and telecommunications news from NVIDIA.
NVIDIA Privacy Policy