The AI-RAN Revolution
- 23 Apr 2026
- Alexander Pabst
Building the Foundation for a Sustainable and Intelligent 6G Future. The past year has been pivotal in shaping the vision for 6G.
Looking back on the progress made in 2025, the 3GPP focused on a set of core priorities: revenue growth; network efficiency; sustainability; and simplification. These are not just aspirations, they are the guiding principles in developing a next-generation network that will support exciting new services, from ubiquitous connectivity to non-terrestrial networks (NTNs) to immersive Extended Reality (XR) experiences, sophisticated sensing applications and the pervasive integration of AI.
At the heart of this transformation lies the AI-RAN, a symbiotic relationship between AI and Radio Access Networks. This isn’t simply about using AI in the RAN, it is about a reinforcing loop that unlocks technical advancements and a compelling business case. Three key aspects have emerged: AI on the RAN, offering new AI and computing services; AI for the RAN, enhancing performance, efficiency and operation; and AI and the RAN, shared infrastructure enabling network functions and third-party AI services
This cycle is remarkably powerful. Hosting AI services at the edge fuels low-latency, compute-intensive applications generating invaluable user data. Collocating RAN functions and third-party AI on the same servers maximises usage and yields rich telemetry. This data fuels the training and deployment of AI models optimising scheduling, reducing latency, enhancing reliability and minimising energy consumption. This reduces friction and cost for AI applications, justifying further investment in edge computing and intelligent RAN control.
Addressing the uplink challenge
However, this AI-RAN vision is not without challenges.
Maintaining robust uplink performance at higher frequencies is a hurdle. The shift from 3.5GHz to 7GHz introduces a 4.5dB increase in Free Space Path Loss (FSPL), compounded by increased attenuation through building walls. Increased device output power is limited by Specific Absorption Rate (SAR) regulations.
The answer lies in intelligent optimisation. Rohde & Schwarz focused on two strategies: reducing communication protocol overhead; and increasing base station receiver sensitivity.
Our MWC Barcelona 2025 demonstration in collaboration with Qualcomm Technologies featured AI/ML-based Channel State Information (CSI) feedback compression, showcasing a “two-sided model” where AI-optimisation is happening in parallel at the device (UE) and the infrastructure (gNB). This improved throughput and validated the ease of testing of such complex models for the first time.
The 3GPP recognised the potential and adopted two-sided models and CSI feedback compression as a work item within Release-20, initiating crucial standardisation efforts for interoperability, focusing on model handling, lifecycle management and ease of testing, ensuring products from different vendors can seamlessly interact.
Enhancing receiver sensitivity
Complementing CSI feedback compression, Rohde & Schwarz made significant strides in link budget for the uplink. A recent collaboration with Nokia Bell Labs demonstrated AI/ML-based base station receivers employing Digital Post Distortion (DPoD) to recover distorted uplink signals.
DPoD boosts the link budget, preserving 5G-like coverage footprints, reducing the need for dense site deployments and yielding substantial cost savings. It also reduces device complexity and power consumption. As XR devices transition from bulky headsets to stylish frames with cameras, audio, video projection and haptic sensing, battery life and distributed compute become paramount for enabling immersive experiences, particularly in XR.
The AI-RAN Alliance made DPoD an official work item in November 2025.
A new dimension for 6G
Beyond core network performance, AI drives new 6G application areas. Integrated Sensing and Communication (ISAC), using mobile networks for object detection, is rapidly gaining traction. The possibilities are vast and the first networks are already in place for tracking vessels in ports.
Other applications defined by the 3GPP include smart traffic management, topographical mapping, healthcare sensing and drone detection.
The 3GPP’s guiding principle of deploying AI where it delivers significant gains ensures a pragmatic approach. Continued industry collaboration, reliable test equipment and trustworthy training data are paramount for navigating interoperability in this continuously evolving landscape.
Alexander Pabst is Vice President of Wireless Communications at Rohde & Schwarz and member of the Board at Global Certification Forum (GCF).
Author
