Conference
AI-RAN’s Reality Check: Smarter Network or Expensive GPU Upgrade?
Date
Wed, 3 Mar
Time
3:45 pm - 4:45 pm CET
Location
Turing Stage, Hall 6
Loading…
Session Description
AI-RAN promises to make the radio access network one of the smartest parts of the telecoms estate, using AI to optimise radio resources, predict congestion, automate fault resolution and dynamically assign capacity to different industrial applications. The early results are compelling: the first commercial AI-native RAN platforms have already shown more than 20% spectral efficiency gains, with ambitions to exceed 100% by 2028. Operators are now running carrier-grade RAN and AI workloads on shared accelerated infrastructure, while AI-RAN increasingly overlaps with autonomous and agentic networks, as multiple AI agents work together to identify faults and coordinate troubleshooting.
The market is following, forecast to grow from US$3.81 billion in 2026 to around US$37.19 billion by 2035. Yet energy is both AI-RAN's best use case and its biggest risk, with the RAN estimated to account for 70–80% of a mobile operator's network energy consumption. Expensive hardware, legacy systems, siloed data, security concerns, rising power bills and unproven profits all stand in the way of successful implementation.