The impact of generative and agentic AI on tech stacks and IT infrastructure is far-reaching, and radio access networking that supports 5G and eventually 6G mobile connectivity also stands to benefit tremendously. The potential to infuse intelligence to convert static, rules-based processing to dynamic, real-time optimization is not a new concept, given the widespread adoption of AIOps within enterprise network deployments. However, more modern applications of AI have enormous potential to enable massive, public mobile networks to self-optimize based on traffic patterns, interference, and even user behavior.
It is no surprise that NVIDIA is leading the charge in AI RAN investigations, demonstrating how AI and machine learning can improve RAN performance and unlock transformative applications at network edges. The company is engaged with 150 operators across the world, which puts it in an enviable position, but other infrastructure providers are also making substantive contributions to AI RAN.
In this piece, I will highlight what NVIDIA and other infrastructure providers are doing to advance AI RAN through consortium and individual engineering efforts, unpack the benefits to mobile network operators, and share how I expect AI RAN to mature over time.
Consortium Efforts
Consortia are vital in bringing technology companies together to codify standards and advance solution development efforts through working groups. The AI-RAN Alliance is no different, and it boasts a membership that includes infrastructure providers Cisco, Dell Technologies, Ericsson, Hewlett-Packard Enterprise, Nokia, NVIDIA, Qualcomm, and Samsung, and operators including Boost Mobile, LG U+, SK Telecom, Softbank, T-Mobile, and others.
The AI-RAN Alliance’s mission is to enhance RAN performance, improve energy efficiency, and increase automation as a precursor to 3GPP 6G standardization. By all measures, it is an audacious endeavor, having the potential to supercharge RAN infrastructure and birth new use cases and applications that could finally move mobility from access and “dumb pipes” to highly automated, smarter networks. Over the past two Mobile World Congress events in Barcelona, AI-RAN Alliance demonstrations point to the transformative potential.
The AI RAN Infrastructure Landscape
Unsurprisingly, NVIDIA is stepping into RAN infrastructure, a segment traditionally dominated by telecom giants like Ericsson, Nokia, and Huawei. The cellular infrastructure incumbents have its own AI-driven network initiatives. Still, NVIDIA’s strategy differs in many ways, buoyed by its overall leadership in facilitating hyperscaler and enterprise generative and agentic AI applications and workloads. NVIDIA’s early lead lies in its ability to bring AI-native computing power to RAN without the baggage of legacy telecom infrastructure. By focusing on a software-defined, GPU-accelerated approach, NVIDIA provides an alternative that is highly adaptable and scalable. Unlike traditional vendors that have spent years optimizing legacy hardware for radio access networks in closed, vendor lock-in ecosystems, NVIDIA is approaching RAN with an AI-first and multi-vendor mindset, developing solutions that are inherently optimized for intelligent network management.
Ericsson has pioneered AI-powered network automation, particularly in closed-loop optimization for 5G networks. However, its AI solutions have typically been more integrated with its RAN infrastructure, limiting interoperability, even given its embrace of more open architectures. Nokia, with its AirScale Cloud RAN, has also taken measurable steps toward AI-enhanced functions, and I also like what the company is doing with its autonomous networks platform – something that I wrote about earlier this year. However, Ericsson’s and Nokia’s involvement in the AI-RAN Alliance could further its AI RAN ambitions and capabilities.
Huawei is heavily investing in AI RAN, but the beleaguered Chinese infrastructure and device giant continues to face geopolitical headwinds that make widespread adoption uncertain. This likely limits its opportunities to capitalize on AI RAN, particularly in Western markets, given the rip and replacement of its core and RAN infrastructure in the rural areas of the United States, the United Kingdom, and other regions globally.
Finally, Opanga Networks may not be as well-known, but it has been focused on machine learning and AI RAN optimization since its inception in 2016. Its RAIN platform, launched in 2023, is deployed globally, enjoying widespread adoption given its hardware, architectural, and vendor-agnostic design. Its platform focuses on surfacing network intelligence, optimization, and energy reduction, all without the need for probes or data injection. This makes it easy to deploy and scale. Furthermore, Opanga claims that based on real-world deployments, it can reduce RAN energy consumption by 40% and improve network performance by 50%.
Longer Term Implications
The telecom sector has been typically slow to embrace bleeding-edge technologies due to legacy vendor ecosystem lock-in and decade-long “G” upgrade cycles. However, NVIDIA and other infrastructure providers are innovating rapidly in the AI RAN space. Consequently, the resulting innovation demonstrated by the AI-RAN Alliance and competitive offerings has great promise in convincing mobile network operators to reevaluate their approaches, given the potential outcomes of more sustainable and highly performant mobile network connectivity.
Tech pundits, including myself, believe that AI will become a fundamental underlying basis for future 6G networks, and these AI RAN efforts could accelerate the development and deployment of next-generation mobile connectivity.
