It’s been a busy few weeks for GPU and AI company Nvidia. In mid-May, the company held the Taipei edition of its GTC user event and then followed that up with CEO Jensen Huang keynoting COMPUTEX 2025. In late May, the company reported better-than-expected results, announcing that its fiscal first-quarter revenue rose 12% from the prior quarter to $44.1 billion— which is a 69% jump compared to the year-earlier quarter.
The AI chipmaker also launched its Aerial RAN Computer-Compact (ARC-Compact). With it, Nvidia aims to bring advanced AI capabilities and software-defined principles to wireless networks. Given that the world is well into the 5G deployment cycle, this product is designed to be fully 6G-ready, enabling seamless integration with 5G investments made between now and the advent of 6G, and providing a smooth, future-proof upgrade path to AI-native wireless networks of the next generation.
Achieving AI-native wireless networks
Nvidia’s key objectives for this effort are to maximize RAN infrastructure utilization (traditional networks average a low 30% to 35%), use AI to rewrite the air interface, and enhance performance and efficiency through radio signal processing. The longer-term goal is to seamlessly process AI traffic, including inference from generative AI, at the network edge, to create new monetization opportunities for service providers.
Nvidia’s plan to make an AI-native network edge a reality is built on four principles:
- All workloads—RAN (radio access network) or AI—must be software-defined and completely decoupled from the underlying hardware, adhering to cloud-native principles for scalability.
- The infrastructure must be AI-native and designed with acceleration in mind to achieve high performance and efficiency, not just for vRAN but also for new AI models and applications.
- A consistent, homogeneous architecture across the entire wireless and centralized network, running core network functions and radio workloads on the same architecture. This simplifies deployments of any size.
- The infrastructure must serve multiple purposes by running on commercial off-the-shelf (COTS) servers, capable of servicing a variety of workloads in addition to RAN.
“RAN platforms of today are very efficient at delivering one thing but cannot handle the future AI applications,” said Soma Velayutham, general manager of Nvidia’s AI and telecom business, in an analyst briefing. “ARC-Compact is a game changer because it is very efficient in handling today’s application and anchor tenant —vRAN— and is well established as the best platform for AI, now and in the future. ARC-Compact is an excellent way for operators to leverage CUDA acceleration, our compact platform delivering both high-efficiency vRAN and advanced AI capability in a single compact platform that is suitable for varying network and site conditions.”
How ARC-Compact meets key requirements for distributed AI-RAN
When Nvidia initially launched ARC-1 (Aerial RAN Computer-1) based on Grace Hopper 200 and Grace Blackwell systems, cellular providers said they needed a smaller, more power-optimized solution for cell sites. This led the company to develop ARC-Compact, specifically designed for distributed AI-RAN scenarios where power efficiency and a small form factor are crucial.
Key features of ARC-Compact include:
- Energy Efficiency: Utilizing the L4 GPU (72-watt power footprint) and an energy-efficient ARM CPU, ARC-Compact aims for a total system power comparable to custom baseband unit (BBU) solutions currently in use.
- 5G vRAN support: It fully supports 5G TDD, FDD, massive MIMO, and all O-RAN splits (inline and lookaside architectures) using Nvidia’s Aerial L1+ libraries and full stack components.
- AI-native capabilities: The L4 GPU enables the execution of AI for RAN algorithms, neural networks, and agile AI applications such as video processing, which are typically not possible on custom BBUs.
- Software upgradeability: Consistent with the homogeneous architecture principle, the same software runs on both cell sites and aggregated sites, allowing for future upgrades, including to 6G.
Velayutham emphasized the power of Nvidia’s homogeneous platform, likening it to the iOS for iPhone. The CUDA and DOCA operating systems abstract the underlying hardware (ARC-Compact, ARC-1, discrete GPUs, DPUs) from the applications. This means that vRAN and AI application developers can write their software once, and it will run seamlessly across different Nvidia hardware configurations, which future-proofs deployments.
Power-efficient and cost-competitive
There has been some skepticism around whether the GPU-powered vRAN can match the power and cost efficiency of custom BBUs. Nvidia asserts that they have crossed a tipping point with ARC-Compact, achieving comparable or even better energy efficiency per watt. The company didn’t disclose pricing details, but the L4 GPU is relatively inexpensive (sub-$2,000), suggesting a competitive total system cost (estimated to be sub-$10,000).
The path to AI-native RAN and 6G
Nvidia envisions the transition to AI-native RAN as a multi-step process:
- Software-defined RAN: Moving RAN workloads to a software-defined architecture.
- Performance baseline: Ensuring current performance is comparable to traditional architectures.
- AI integration: Building on this foundation to integrate AI for RAN algorithms for spectral efficiency gains.
Nvidia believes AI is ideally suited for radio signal processing, as traditional mathematical models from the 1950s and 60s are often static and not optimized for dynamic wireless conditions. AI-driven neural networks, on the other hand, can learn individual site conditions and adapt, resulting in significant throughput improvements and spectral efficiency gains. This is crucial given the hundreds of billions of dollars providers spend on spectrum acquisition. Nvidia has said it aims for an order-of-magnitude gain in spectral efficiency within the next two years, potentially a 40x improvement from the last decade.
To make this possible, Nvidia tools, including the Sionna and Aerial AI Radio Frameworks, support rapid development and training of AI-native algorithms. The “Aerial Omniverse Digital Twin” enables simulation and fine-tuning of algorithms before deployment, mirroring the approach used in autonomous driving, another area of focus for Nvidia.
Market approach and benefits
Nvidia is working closely with OEM and ODM vendors to build servers and with Network Equipment Providers (NEP), including Nokia and Samsung, to run their vRAN on Nvidia platforms.
Key benefits of ARC-Compact
Nvidia is touting several benefits it says ARC-Compact will deliver to the wireless industry, including:
- Energy efficiency and right-sizing: ARC-Compact provides high Gbps/watt efficiency and is compatible with typical cell site power and space constraints.
- AI-native radio: Uses AI algorithms to achieve massive spectral efficiency gains.
- Homogeneous software architecture: Supports deployment across any platform, abstracting hardware from software.
- Multi-purpose COTS servers: These run not only vRAN and AI applications but also other telco workloads, including UPF (user plane function), packet core elements, and virtualized cell site routers, leveraging the ARM ecosystem.
- Future-proofing for 6G: The platform’s software-defined, hardware-agnostic nature positions operators for a seamless transition to future 6G networks, with support for new spectrum, AI-native algorithms, and ultra-massive MIMO (multiple input, multiple output).
Source:: Network World