Ericsson has launched a new software subscription called AI in RAN, bringing telco-grade artificial intelligence models into its radio access network (RAN) portfolio for 5G networks.
The software is engineered to run AI capabilities inside existing baseband and radio equipment, allowing operators to add intelligence without deploying new hardware. Ericsson says the first AI in RAN features will be available in the second quarter of 2026, with additional functionality rolling out later in the year.
Software-based AI for 5G networks
AI in RAN is designed to integrate with Ericsson 5G Advanced across both purpose-built RAN and Cloud RAN platforms. It applies AI models across multiple radio network domains to improve performance, reduce energy consumption, and automate operations.
The solution leverages Ericsson Silicon for on-radio AI inference and the latest generation of RAN Compute for distributed processing. The Cloud RAN software supports portability, enabling deployment on compatible partner platforms.
This launch builds on previous Cloud RAN collaborations between AT&T, Ericsson, and Intel. In March 2026 the partners demonstrated AI-native Link Adaptation on a Cloud RAN stack running on Intel Xeon 6 SoC, reporting up to 20% higher throughput compared with traditional rule-based link adaptation.
Ericsson emphasizes that its models are optimized for microsecond-level inference, meeting the low-latency demands of radio network functions. The models were developed using domain expertise, trained on high-quality datasets, and designed to adapt to changing RAN conditions.
Initial capabilities include an AI-native scheduler for link adaptation, AI-powered macro positioning, and AI-managed beamforming. Additional features cover multi-layer coordination, performance management event schema, and enhanced observability tailored for AI in RAN.
Link adaptation is the process the network uses to adjust transmission parameters according to current radio conditions. Ericsson’s AI-native implementation relies on real-time models that react to shifting signal quality, interference, and other dynamic factors to improve throughput and reliability.
Trials and operator deployments
Ericsson reports more than 15 trials and deployments worldwide. The company says the technology has yielded up to 20% higher downlink throughput and as much as 10% better spectral efficiency in tested scenarios.
Other reported benefits include support for up to twice as many high-traffic users, coverage prediction accuracy in the 90%–95% range, and up to five times improved user-positioning precision.
Ericsson also tested AI-based coverage prediction with Optus on 5G Standalone networks. That model predicts whether a device served on one frequency layer is also in coverage on another layer—information that helps the network make better decisions.
The company named SoftBank, Bell, SK Telecom, and Rogers among operator partners working with the software. These operators highlighted improvements in network optimization, spectrum efficiency, energy consumption, and AI-enabled services.
SoftBank noted the software’s potential to support low-latency services and physical AI use cases that require tight coordination between network and compute resources.
SK Telecom said its collaboration with Ericsson blends research, field validation, and software development as part of preparations for AI-native 6G networks.
Bell and Rogers emphasized performance gains and energy savings. Rogers specifically said it plans to use AI in RAN for real-time network optimization and reduced power consumption.
Mobile Experts principal analyst Joe Madden commented that the software upgrade provides operators a way to increase capacity and improve network observability using existing 5G assets, and noted the potential for more accurate location-based services.
Ericsson has been integrating AI into its network products since the 4G era. In 2021 the company introduced AI-ready acceleration in RAN Compute, and in February 2026 it announced Neural Network Accelerators in its Massive MIMO radios, citing a tenfold increase in on-radio AI inference capability.
(Photo by Scott Rodgerson)
Interested in how IoT is transforming telecoms and connectivity? Consider attending the IoT Tech Expo events in Amsterdam, California, and London to learn how 5G, edge computing, and IoT innovations are shaping future networks and services. The expos are part of TechEx and are co-located with other leading technology conferences.
Telecoms Tech News is produced by TechForge Media. Explore additional enterprise technology events and webinars offered by the publisher.