AI-RAN Implementation Strategies for 5G and 6G Networks

Why It Matters

AI‑enabled RAN delivers immediate cost and performance gains for 5G while laying the architectural foundation needed for fully autonomous, AI‑native 6G networks, giving operators a competitive edge in the next generation of mobile services.

Key Takeaways

• AI RAN already delivering 25% energy savings in 5G deployments.
• Real‑time AI inference runs on CPUs with 60‑80 µs latency.
• GPUs remain essential for AI training, not cost‑effective for inference.
• 6G aims for AI‑native RAN, integrating intelligence into air interface.
• Hybrid CPU‑GPU edge platforms will support low‑latency AI workloads.

Summary

The interview with Vihang Campbell, CTO of Rakuten Symphony’s RAN unit, explores how artificial‑intelligence‑driven radio access networks (AI RAN) are being rolled out in today’s 5G and what they could look like in future 6G deployments. Campbell highlights that AI RAN is no longer experimental; Rakuten Mobile has integrated third‑party r‑apps on the O‑RAN RIC platform, achieving roughly 25% energy savings and measurable gains in uplink robustness, voice quality, and adaptive modulation error rates.

Key technical insights include a clear split between non‑real‑time AI (handled in the cloud) and real‑time AI at the edge. Real‑time inference runs on modern CPUs with 60‑80 µs latency, while GPUs are reserved for training workloads due to cost considerations. Demonstrations showed error‑rate cuts from 20% to 10% and voice‑quality jumps from poor to good, underscoring the tangible performance uplift AI can deliver.

Campbell stresses that 6G will move beyond “AI‑assisted” to “AI‑native” RAN, embedding intelligence directly into the air‑interface protocol and baseband software. He calls for bold actions from 3GPP and the O‑RAN Alliance to decouple software from hardware, adopt COTS platforms, and enable option 7.2x splits for third‑party radio units. The vision includes intent‑based networking and service‑based management that auto‑configure slices with minimal human intervention.

For operators, the message is clear: AI RAN can slash operational expenditures, improve user experience, and future‑proof networks for the software‑centric, ultra‑low‑latency demands of 6G. A hybrid edge architecture—CPU‑first with optional GPU acceleration—will be essential to support emerging workloads such as XR, real‑time gaming, and sensing applications.

Original Description

Rakuten Symphony’s CTO, Vihang Kamble, explains how AI RAN is moving beyond experimental phases in 5G networks, with practical implementations cutting energy use by 25% while significantly improving voice quality and error rates. He advocates for CPU-based inference rather than GPU deployment for cost-effective 5G implementations, while planning ahead for 6G networks that feature deeper AI integration, autonomous operations, and intent-based networking capabilities that could eliminate manual parameter tuning.

Featuring: Vihang Kamble, CTO, RAN BU, Rakuten Symphony

Recorded June 2026

#AIRAN #5G #6G #OpenRAN #NetworkAutomation #TelecomTV