What's Happening?
A new survey from Omdia indicates that telecom operators are making tangible progress towards deploying Artificial Intelligence (AI) models directly within the Layer 1/2/3 radio stack of radio base stations, signaling a move towards AI native RAN (Radio
Access Network). A small leading group of 12% of respondents have already deployed or plan to deploy AI models, while nearly a quarter (27%) are conducting live network trials. These AI models, deployed as software updates, aim to enhance RAN system performance, including coverage, energy consumption, uplink performance, and spectral efficiency. Example use cases include digital pre-distortion, AI-managed beamforming, and AI-optimized link-adaptation. The survey highlights that initial deployments are just the beginning, with potential for multiple AI models to run simultaneously to improve overall system performance.
Why It's Important?
The integration of AI into RAN is a critical development for the U.S. telecom industry, promising significant improvements in network efficiency, performance, and user experience. By optimizing radio processes beyond conventional algorithms, AI RAN can lead to better coverage in challenging areas, reduced energy consumption for base stations, and more efficient use of spectrum. This can translate into cost savings for operators and enhanced service quality for consumers and businesses across the U.S. The move towards AI native RAN is also foundational for future 6G architectures, positioning the U.S. telecom sector at the forefront of next-generation wireless technology. The survey's findings suggest a clear direction for investment and innovation in AI accelerators and software-defined RAN.
What's Next?
While initial deployments and trials are underway, the survey also indicates that extensive and deep deployment of AI models in the world's millions of radio base stations will take time, with 27% in lab trials and 30% in the desk research phase. Operators will continue to explore how multiple AI models can run simultaneously and interact to maximize performance gains. The development of AI accelerators, whether embedded in custom RAN silicon or dedicated xPU devices, will be critical for scaling AI RAN. The industry will also focus on deploying models tailored for different network environments, such as rural cell edges versus dense urban locations. The long-term goal remains a fully AI native, AI-controlled network, which is still a research objective.
Beyond the Headlines
The shift towards AI RAN represents a fundamental change in how wireless networks are designed, operated, and monetized. Beyond immediate performance gains, this evolution has broader implications for the telecommunications ecosystem. It could foster new partnerships between AI developers and telecom equipment providers, driving innovation in both software and hardware. The emphasis on energy efficiency through AI optimization also aligns with global sustainability goals, potentially reducing the carbon footprint of extensive network infrastructure. Furthermore, the ability to deploy AI models trained for specific network conditions and use cases suggests a future of highly adaptive and intelligent networks, capable of dynamically responding to changing demands and environments. This could lead to entirely new services and applications that leverage the enhanced capabilities of AI-driven wireless communication.













