Nokia and NVIDIA Edge AI-RAN Closer to Reality

October 2026

Telecoms & Connectivity

Ardit Ballhysa

Senior Research Analyst

In October 2025, NVIDIA and Nokia announced a partnership, with NVIDIA taking a 2.9% stake in Nokia worth $1 billion, and Nokia agreeing to move its 5G and 6G Radio Access Network (RAN) software to NVIDIA’s Compute Unified Device Architecture (CUDA) platform. Following this, in July 2026, Nokia claimed it had built the industry’s first commercial AI-RAN platform, which it aims to sell to operators throughout 2027.  

On 21st October 2026, Nokia and NVIDIA announced progress on their collaboration, with Nokia announcing that it is applying AI within the 5G signal-processing chain, including channel estimation, beamforming, multi-user Multiple-Input Multiple-Output (MIMO) pairing, and scheduling. Under their partnership, the two companies have established a platform built around enabling three capabilities: AI-led spectral efficiency to extract more capacity from existing spectrum, integrated sensing and communications to help the network understand the physical environment, and localised inference to process intelligence closer to where data is generated. 

The October 2026 announcement also highlighted progress with major operators, including Elisa, BT, Deutsche Telekom (DT), and Orange. Elisa ran voice, data, and video analytics simultaneously on Nokia’s AI-RAN platform in its own 5G network, which showcased how AI-RAN can support radio and AI workloads in a live network environment, with the two companies announcing that they will launch commercial AI-RAN services in 2027. BT, on the hand, will deploy a live-network trial in Birmingham, UK, in 2027, while Orange plans to progress its testing of AI-RAN in both laboratory and real network conditions. 

Furthermore, Nokia is helping DT develop AI-based channel estimation to improve network performance and efficiency. According to Nokia, initial laboratory measurements showed typical uplink throughput improvements of 10–15%, with greater gains under more complex radio conditions. The next stage of this partnership will incorporate additional AI spectral-efficiency algorithms and field validation. 

While these announcements are positive steps towards further AI deployment within the RAN, they represent different stages of technical validation, rather than evidence of strong efficiency gains or commercial opportunity. However, the growing list of infrastructure partnerships, which will provide the deployment foundation for the three capabilities mentioned above, are significant and noteworthy. 

Dell Technologies, Quanta Cloud Technology, and Supermicro support flexible NVIDIA Aerial RAN Computer (ARC) configurations for distributed and centralised deployments running Nokia 5G anyRAN software alongside AI inference workloads. Meanwhile, Red Hat is utilising its cloud-native, Red Hat OpenShift platform to provide a commonplace for AI-RAN technologies to create a foundation for RAN functions and AI workloads on shared infrastructure. 

Operator Readiness

The proponents of AI-RAN claim it will provide major advantages to operators; with the biggest two being significant cost savings, and the potential to generate new revenue streams. This will be particularly relevant over the next five years, as data traffic is continually growing. This growth is due to an increase in connected devices across both consumers and enterprises, which is increasing the capacity needed on networks, and requiring additional spectrum, radio equipment, site upgrades, energy, and backhaul. Conversely, this increase in connected devices also creates avenues for operators to generate potential revenue streams, especially among enterprise customers, where widespread adoption of physical AI will not only require large streams of data, but also secure, ultra-low latency connectivity that operators could offer at a premium. 

While the laboratory measurements with DT showed an improvement in uplink throughput, at a time where operational costs are high and consumers and enterprises are budget constrained, some operators will not be convinced that investments in expensive AI-RAN equipment are justified. Even more so as the measurements were achieved under laboratory conditions, which do not necessarily translate to real-world environments, especially dense, urban environments. Unless these operators see consistent and significant improvements in real-world environments, they will continue to upgrade their networks without incorporating AI-RAN. 

In addition, following the hype vendors placed on Open RAN, some operators are wary of claims that new technology and standards will revolutionise the RAN. Open RAN was touted as an interoperability standard that would provide benefits to operators by enabling multi-vendor deployments of RAN equipment. By making RAN equipment interoperable and reducing vendor lock-in, operators were told that they would have greater flexibility when upgrading RAN equipment, from a larger choice of suppliers; ultimately allowing them faster and more cost-effective upgrade options. The Open RAN movement, however, has not materialised its promised impact. 

Stronger Evidence Is Needed

Convincing operators of the benefits of AI-RAN through claimed theoretical improvements or laboratory tests is not sufficient. Nokia and NVIDIA need to deploy larger-scale, real-world environment deployments across different situations, including urban environments, to highlight efficiency benefits to operators. Following this, they need to provide tangible evidence of any cost savings associated with these deployments. Furthermore, the two companies should conduct deployments to highlight new use cases enabled by their platform, with particular focus on additional revenue streams created from these new use cases.