In sum – what to knowO-RAN foundation – Rakuten Mobile says its software-defined, cloud-native architecture provides the foundation for adding AI intelligence to its network.
AI moves to edge – Rakuten Mobile and Intel are working to enable AI inference within the RAN environment, including at the cell edge.
Japanese carrier Rakuten Mobile is using the software-defined foundation of its Open RAN (O-RAN) network to move toward AI-native automation, with energy efficiency and edge inference emerging as key areas of its strategy.
Rakuten Mobile launched commercial mobile services in Japan in 2019 and has become one of the industry’s highest-profile Open RAN operators.
Last year, Rakuten Mobile had named Cisco, Nokia, and F5 as its lead technology partners for the deployment of its 5G Standalone (SA) network.
In sum – what to know
O-RAN foundation – Rakuten Mobile says its software-defined, cloud-native architecture provides the foundation for adding AI intelligence to its network.
Autonomy with guardrails – The operator has achieved TM Forum Level 4 autonomous validation for RAN energy efficiency.
AI moves to edge – Rakuten Mobile and Intel are working to enable AI inference within the RAN environment, including at the cell edge.
Japanese carrier Rakuten Mobile is using the software-defined foundation of its Open RAN (O-RAN) network to move toward AI-native automation, with energy efficiency and edge inference emerging as key areas of its strategy.
Speaking at the recent RCR Wireless News’ Intelligent RAN Forum, Sudhakar Pandey, head of RAN at Rakuten Mobile, said the operator’s network has grown from its 2019 O-RAN launch to more than 152,000 sites and over 330,000 cells. The initial objective was to break vendor lock-in and establish a software architecture, he said, with O-RAN now providing the foundation for adding AI intelligence to the network.
One example is RAN energy efficiency. Rakuten Mobile recently achieved TM Forum Level 4 autonomous validation for energy efficiency in a live network. Pandey said the initiative has delivered more than 20% energy savings, contributing to the company’s 2030 sustainability target and EBITDA.
Pandey said the benefits were achieved within four to six weeks after deployment, which he attributed to the network’s fully virtualized, cloud-native foundation. He also said Rakuten Mobile is among very few operators globally to have received TM Forum Level 4 validation for energy efficiency.
The operator’s approach to autonomy still includes human oversight. Pandey said autonomy does not mean eliminating human intervention, with people determining what should be done, the risk appetite and the guardrails under which automation operates. Defined criteria can then be managed through closed-loop automation, while human intervention remains necessary when something falls outside those guardrails or when the model encounters something it has not been trained on.
Customer experience is also incorporated into the process. In addition to traditional network KPIs such as availability, accessibility, and mobility, Rakuten Mobile monitors customer-centric measures including latency, packet loss, jitter, voice quality, and call setup time. Pandey said this allows the operator to improve efficiency while keeping user experience intact.
He attributed improvements in customer experience to a combination of site additions and capacity, continuous radio-frequency optimization and software upgrades. Because the network is software-defined, Pandey said the same software can run across different generations of hardware, allowing the Japanese carrier to upgrade both its platform and RAN software.
The operator is also working with Intel to bring AI inference closer to the network. Pandey said Rakuten Mobile’s RAN software is based on Intel’s FlexRAN reference model and that the companies are working with what he referred to in the presentation as Intel’s latest “GNRD” chipset.
According to Pandey, the companies are working to embed inference into the vRAN or O-RAN stack so it can run alongside RAN workloads rather than requiring processing to be sent back to a central data center. He said this can provide advantages for applications where decisions made at the cell edge can improve user experience.
For operators considering AI-native RAN, Pandey emphasized that the transformation is not simply technological. He described it as both a technological and cultural transformation, requiring a shift toward disaggregated, software-defined networks.
Rakuten Mobile’s experience, he said, demonstrates that disaggregation can work at scale and at the required economics. He added the company has optimized costs while maintaining the quality of service offered by other operators in Japan.
Rakuten Mobile launched commercial mobile services in Japan in 2019 and has become one of the industry’s highest-profile Open RAN operators. Last year, Rakuten Mobile had named Cisco, Nokia, and F5 as its lead technology partners for the deployment of its 5G Standalone (SA) network. The Japanese company noted that these alliances will “significantly enhance its network capabilities, simplify operations through AI-driven systems, and drive innovation” across Japan.
