Officials at various U.S. government agencies are expressing concern over the progress of 5G technology, which has not delivered the anticipated enhancements compared to 4G. This sentiment is echoed by executives in the private sector, including those at Nvidia, a leading U.S. chipmaker. Despite the initial optimism surrounding 5G, its capabilities have largely been limited to increased capacity, with the technology not designed to support the rapid advances in artificial intelligence (AI) that emerged shortly after its standardization.
The dominance of European companies Ericsson and Nokia as traditional gatekeepers of the telecommunications network continues to raise concerns. A movement known as open RAN aimed to revolutionize the industry by introducing industry-standard interfaces to replace proprietary ones used in conventional networks. Unfortunately, this initiative has not significantly altered market dynamics. As the Mobile World Congress (MWC) tradeshow unfolds in Barcelona, attention has shifted towards a new approach emphasizing open source solutions for the forthcoming 6G network.
The Department of War has collaborated with the Linux Foundation, a prominent open-source organization, on an initiative dubbed OCUDU. This program aims to integrate open-source code into the next generation of telecommunications. Nvidia is advocating for an “open” and “open source” philosophy in its own 6G project, joining forces with major telecom companies including BT, Deutsche Telekom, SK Telecom, SoftBank, and T-Mobile. Notably, both Ericsson and Nokia are also participating in OCUDU.
In telecommunications, the term “open” is often misinterpreted. While the 3GPP, the organization responsible for evolving cellular standards, claims to operate on open principles, it generates significant revenue through licensing its technologies. This licensing model is seen as incompatible with true open-source ideals. According to Tom Rondeau, who oversees the Department of War’s FutureG network project, the existing frameworks, including open RAN, remain too restrictive, limiting opportunities for smaller companies to innovate.
Vasishta, who leads Nvidia’s telecom activities, agrees that open sourcing could provide developers with greater flexibility across the entire technology stack. He argues that if open source is widely adopted for 6G, it would enable developers to integrate new algorithms seamlessly into the broader platform. Nvidia’s existing open-source reference platform, known as Aerial, exemplifies this potential. It allows developers like DeepSig to incorporate AI-driven waveforms into the Aerial stack.
Nvidia’s vision for AI-driven radio access networks (RAN) emphasizes replacing traditional algorithms with AI to enhance spectral efficiency. However, Aerial currently operates within the constraints of Nvidia’s software platform, CUDA. This limitation poses challenges for broader deployment on general-purpose central processing units (CPUs) from companies like Intel or AMD. Vasishta expressed skepticism about the feasibility of fully addressing hardware dependency through open source alone.
In recent collaborations, Nvidia and Nokia have trialed the integration of their systems, with T-Mobile US participating since last year. This close partnership followed Nvidia’s significant investment of $1 billion in Nokia. Vasishta argues that leveraging graphics processing units (GPUs) for RAN functions can enhance performance and efficiency, suggesting that many Layer 1 tasks are well-suited for GPU execution.
In contrast, Ericsson appears to be adapting its software initially designed for Intel chips to function on Nvidia’s Grace CPU, while offloading specific tasks to Nvidia’s Hopper-branded GPUs. This divergence in approach could influence future developments in AI-driven RAN systems, with each company pursuing different paths.
Despite the potential benefits of integrating GPUs into RAN architecture, concerns remain about the associated costs and energy consumption. Vasishta acknowledges that improvements in performance per watt make GPUs increasingly viable, suggesting a shift towards software-defined RAN platforms on high-performance computing systems is inevitable.
The involvement of Ericsson and Nokia in OCUDU raises questions about their long-term business strategies. Rondeau notes that collaboration with the U.S. government and military offers benefits at a time when public spending on civilian networks is declining. However, embracing open-source 6G could disrupt existing business models and introduce new competitors.
According to Ranny Haiby, the Linux Foundation’s chief technology officer for networking, it is unlikely that Ericsson and Nokia will immediately abandon their existing proprietary technologies. He anticipates a gradual transition towards greater compatibility with open-source initiatives, although the timeline for such changes remains uncertain.
The evolving nature of AI poses additional challenges for standardization processes. Ultan Mulligan, chief services officer at the European Telecommunications Standards Institute (ETSI), emphasizes that the rapid evolution of AI may create difficulties in defining standards for 6G. The need to remain adaptable is crucial, as premature decisions could exclude valuable capabilities from future networks.
As the telecommunications landscape shifts towards more open-source solutions, traditional players face a turbulent period. The future of 6G may hinge on balancing innovation with established business practices, and the outcome of these developments will have far-reaching implications for the entire industry.
