en.Wedoany.com Reported - Recently, Swedish telecommunications equipment provider Ericsson launched the AI in RAN software subscription, introducing carrier-grade AI models directly into the baseband and radio units of the Radio Access Network for communication operators. This solution can be used to enhance 5G network performance, spectrum efficiency, energy management, and automated operations capabilities without adding new hardware, while supporting operators in their evolution towards an AI-native RAN. The initial features were made available in the second quarter of 2026, with subsequent enhancements to be rolled out within the year.
The focus of AI in RAN is not simply connecting external large models to the network management system, but embedding AI models into the real-time decision-making positions of the wireless network. 5G base stations need to complete link adaptation, beam management, user positioning, multi-layer coordination, and resource scheduling within extremely short timeframes. Traditional rules and manual parameter optimization struggle to cope with high traffic, complex scenarios, and dynamic loads. With this software subscription, Ericsson places AI inference capabilities within baseband computing and radio units, enabling the network to perform more granular adjustments based on actual traffic, radio environment, and user distribution. Official information indicates that its carrier-grade AI models are designed for microsecond-level low-latency inference and emphasize reliability and robustness in diverse wireless environments.
This product also reflects that 5G construction is entering a phase of "intelligentization of existing networks." Many operators have completed large-scale 5G investments but still face challenges such as network capacity growth, energy consumption pressure, operational complexity, and insufficient revenue conversion. If AI in RAN can enhance existing 5G network capabilities through software upgrades, it allows operators to unlock more capacity and efficiency without large-scale equipment replacement. Data released by Ericsson states that in over 15 deployments and trials globally, the solution achieved up to a 20% increase in downlink throughput, up to a 10% improvement in spectrum efficiency, and can support up to twice the number of high-traffic users, while coverage prediction accuracy reaches 90% to 95%.
From a functional design perspective, the initial capabilities include AI-native link adaptation scheduling, AI-driven macro cell positioning, AI-managed beamforming, multi-layer coordination, performance management event pattern files, and AI in RAN enhanced observability. These functions address long-standing engineering challenges in wireless networks: maintaining a stable user experience amidst user mobility, obstructions, interference, sudden load changes, and multi-band coordination. In the past, operators relied on manual optimization, periodic parameter adjustments, and traditional algorithms to handle these issues. With AI models entering the RAN, network optimization can shift from post-analysis to near real-time closed-loop control.
The industry significance of Ericsson's latest update lies in AI beginning to move from customer service, marketing, and backend operations tools in the telecommunications industry into the core control chain of base stations and wireless networks. Operators such as SoftBank, Bell, SK Telecom, and Rogers have expressed interest in real-time optimization, energy efficiency, automation, and network foundations oriented towards AI services. As AI applications, AR terminals, industrial connections, the Internet of Vehicles, and low-latency services increase, wireless networks need to carry more uplink traffic, higher reliability, and more stable positioning capabilities. Only after RAN-side AI capabilities mature can operators conditionally turn network slicing, edge computing, differentiated connectivity, and enterprise-grade assurance services into more billable products.
Going forward, the procurement pace of operators and actual network performance remain to be observed. AI in RAN adopts a software subscription model, and its commercial success depends on its stable operation across different frequency bands, varying urban densities, different equipment generations, and Cloud RAN environments, as well as on whether operators are willing to pay separately for network performance and automation improvements. For Ericsson, such products help extend 5G equipment sales into long-term software revenue; for operators, the key lies in whether the capacity, energy efficiency, and experience improvements brought by software upgrades can be translated into quantifiable benefits. As 5G Advanced continues to advance, embedding AI into the Radio Access Network will become a new focal point of competition among telecommunications equipment vendors.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
