Rethinking the Radio Access Network with AI and Silicon
Nokia is pivoting its radio access network (RAN) strategy, deeply integrating with Nvidia's computing platforms, specifically their Graphics Processing Units (GPUs). This collaboration moves Nokia beyond its current approach of using ARM-based chipsets to accelerate lower-layer RAN functions. The company is now porting its core RAN software, including the computationally intensive Layer 1 processing, to run directly on Nvidia's GPU architecture. This strategic shift aims to harness the power of GPUs for AI-driven network operations and lays the groundwork for future 6G technologies, which are envisioned as inherently AI-native.
The immediate implication of this partnership is the potential to transform how telecom networks function. By leveraging Nvidia's GPUs, Nokia's base stations can simultaneously handle commercial 5G traffic, artificial intelligence queries, and real-time applications like video captioning. This integration allows telecom operators to tap into currently unused computing power within their radio networks and repurpose it for external AI tasks. This also enables the execution of AI inference computations at the network edge, allowing base stations to provide AI services while continuing their primary communication duties.
Read More: Andhra Pradesh attracts Rs 5,001 crore LG Electronics investment for new plant
A Deeper Embrace of GPUs for AI-RAN
Nokia's commitment involves accelerating the development of its 5G and 6G RAN software specifically for Nvidia's platform, including its CUDA. This represents an evolution from its previous strategy, which focused on plug-in accelerators and ARM-based solutions as a bridge to explore the potential of AI-RAN. The companies have announced a long-term collaboration, with future Nokia base stations slated to incorporate Nvidia's ARC platform, combining Grace CPUs with Blackwell GPUs. This move positions AI as a central element in Nokia's 6G roadmap, where every network component is expected to possess AI integration capabilities.
The partnership extends to virtual RAN (vRAN) initiatives. Nokia plans to utilize Nvidia's Grace CPU Superchip for vRAN functions at Layer 2 and above, integrating it with Nokia's RAN software and hardware accelerators. This aspect of the deal aims to provide more choices within the vRAN ecosystem. While many companies are forging alliances with Nvidia, Nokia's deep integration signifies a significant bet on the GPU as a core component for future network evolution.
Read More: Dropbox uses AI to make searching files faster for companies
"Nokia intends to accelerate development of Nokia’s 5G & 6G RAN software to run on NVIDIA’s architecture." - Nokia
"Companies such as BT, Elisa, NTT DOCOMO, and Vodafone Group are already using Nokia’s AI-RAN technology to improve their networks." - EE Times
Transforming Network Economics and Capabilities
The economic underpinnings of this transition are significant. Historically, cellular networks have relied on costly, rigid deployment cycles and specialized silicon for radio signal processing. The integration with Nvidia's GPU platforms offers a more flexible and potentially cost-effective approach. By enabling simultaneous processing of diverse workloads, from standard communication to advanced AI, on a single server, telecom providers can optimize resource utilization.
Nokia is actively demonstrating these capabilities through working tests of its software on Nvidia's GPU platforms, reportedly with major telecom operators. This practical validation underscores the company's vision of a transformed telecommunications landscape driven by AI and powerful silicon. The goal is to unlock new revenue streams and operational efficiencies by making the RAN a more versatile computing platform.
The reliance on GPUs for core RAN functions, particularly Layer 1, represents a departure from traditional approaches and signals a strong belief in the scalability and processing power offered by these components for demanding, real-time network tasks.
Read More: BP Appoints New Boss Meg O’Neill for £11.7m Amid Falling Profits
The broader implications for the vRAN ecosystem are also noteworthy. By working with Nvidia's CPUs and GPUs, Nokia aims to foster greater diversity and innovation within the virtualized RAN space. This could lead to more open and adaptable network infrastructures, moving away from proprietary solutions.
Nokia's move involves porting its RAN software, including Layer 1, onto Nvidia's GPU platform.
This collaboration aims to accelerate the development of 5G and 6G RAN software for Nvidia's architecture.
Telecom operators can potentially leverage their RAN assets for external AI tasks.
Future Nokia base stations will integrate Nvidia's ARC platform, featuring Grace CPUs and Blackwell GPUs.
The partnership also supports Nokia's virtual RAN (vRAN) strategy with Nvidia's CPUs.
Background: The Evolving RAN Landscape
The Radio Access Network (RAN) is the part of a telecommunications system that connects a user device, like a smartphone, to the core network. Traditionally, this involved specialized hardware components for radio signal processing. However, the industry has been shifting towards virtualizing these functions, allowing them to run on standard servers, often referred to as virtual RAN (vRAN) or Open RAN.
Artificial Intelligence (AI) is increasingly being integrated into network operations to optimize performance, manage resources, and enable new services. AI-RAN refers to the integration of AI capabilities directly within the RAN. This includes using AI for tasks like signal processing, network optimization, and enabling edge computing applications.
Read More: AI development faster than rules, experts worry about control
Nvidia is a dominant player in the field of GPUs, which are highly effective at parallel processing tasks, making them well-suited for AI workloads and increasingly, for traditional computing tasks previously handled by CPUs. Their involvement in the telecom sector signifies a growing recognition of the importance of advanced silicon in building next-generation networks.