• The race is on to make the radio access network (RAN) more efficient and customer friendly through the use of AI
• Ericsson is embedding AI tools directly into its RAN platforms to enable greater optimisation, automation and energy efficiency
• Customers such as SoftBank Corp, SK Telecom and Bell Canada are positive about the results

Ericsson has taken the next step in its AI tech strategy by embedding AI tools in its radio access network (RAN) platforms and offering them to mobile network operators, some of which have been involved in trials, as a software upgrade on a subscription model. 

The Swedish vendor calls this AI in RAN, noting that this approach “brings telco-grade AI models into basebands and radios to boost efficiency, performance and energy savings”. From the perspective of the three pillars proposed by the AI-RAN Alliance (of which Ericsson was a founding member) at its launch in February 2024, this aligns with the AI-for-RAN approach (the use of AI tools to improve the performance and efficiency of radio access networks). 

Whatever it’s called and whatever it comprises, it is touching on one of the most critical areas of mobile infrastructure development right now – the introduction of increasingly energy-efficient, AI-enabled automated network optimisation. The vendor claims AI in RAN “delivers up to 20% higher downlink throughput and up to 10% better spectral efficiency. It also supports up to 2x more high-traffic users, 90%-95% coverage prediction accuracy, and achieves up to 5x greater user-positioning precision.”

The vendor, which earlier this year unveiled “AI-ready radios”, says its AI-in-RAN functionality includes: “Telco-grade AI models designed to run in real time” within the RAN, which are “designed for ultra-low latency inference at the microsecond level”, making them applicable in physical AI/robotics and other use cases where such low latency is required; continuous learning software tools; and “agentic AI support to enable advanced RAN automation and network operations”. 

Examples of the applications on offer are AI-native Scheduler for Link Adaptation, AI-powered Macro Positioning, AI-managed Beamforming, AI-powered Multi-layer Coordination, Performance Management Event Schema Files, and Augmented Observability for AI in RAN. The initial AI-in-RAN features are on course to be available by the end of June.

It’s a bit unclear, though, what mobile operators will need to have deployed in order to run these AI models and applications. The vendor notes that “AI in RAN works with Ericsson 5G-Advanced across both purpose-built and Cloud RAN platforms to enable new AI-driven services… [it is] designed to use the right AI model in the right part of the radio network, powered by Ericsson Silicon for energy-efficient AI inference in radios and the latest generation of RAN Compute, while the portability of Cloud RAN software enables AI capabilities to be deployed across partner platforms.” 

This suggests that a 5G standalone (5G SA) core platform, as well as 5G Advanced software, would be a pre-requisite for the use of AI in RAN functionality. As you’d expect, Ericsson would like all of this to run on its own hardware but there is also the option in cloud RAN deployments to run AI-in-RAN tools on “partner platforms”, though what these might be, or whether running the tools on third-party systems would mean the tools are not fully optimised, is unclear. (We have submitted questions to Ericsson on these matters, as well as to ask how the subscription model works for mobile operators.)

Whatever the answers, Ericsson-friendly operators, 15 of which have either trialled or already deployed the AI-in-RAN solution, are bullish.

Teruyuki Oya, chief network officer (CNO) at Japanese operator SoftBank Corp, stated: “Ericsson’s AI-in-RAN software marks an important step in bringing AI deeper into the radio access network. By enabling real-time optimisation of radio performance, spectrum efficiency and user experience, it helps us turn AI innovation into practical value on live networks. We also see strong potential in how this foundation can support emerging AI-driven services, including physical AI scenarios that depend on low-latency, highly reliable connectivity and intelligent coordination between network and compute resources.”

Bruce Dean, senior VP of network technology and operations at Bell Canada, noted: “At Bell, we’re continuously evolving our network to meet growing demand for high-performance, AI-driven services. Integrating AI directly into the RAN is an important step in making networks more intelligent and efficient. Working with partners like Ericsson, we’re bringing these capabilities into our network to enhance performance, improve energy efficiency and deliver a better experience for our customers.”

Yu Takki, head of SK Telecom’s Network Technology Office, stated: “Through our collaboration with Ericsson, SK Telecom is advancing AI-RAN to enhance network performance and energy efficiency while supporting more intelligent and automated operations. By combining research, real-world validation and software innovation, we aim to strengthen our leadership in AI-powered network evolution and help lay the foundation for AI-native 6G.”

- Ray Le Maistre, Editorial Director, TelecomTV