Techno-Economic Assessment and Return on Investment of Open RAN and AI-RAN Architectures in 6G Deployments

High CapEx and flat ARPU in 5G networks create serious financial risks for 6G rollouts, where the Radio Access Network (RAN) still accounts for 60% to 70% of Total Cost of Ownership (TCO). This paper models a 10-year TCO and Return on Investment (ROI) comparison between COTS-based Open RAN (Intel FlexRAN VRB1, 300 W) and GPU-based AI-RAN (NVIDIA GH200 Grace Hopper, 1,200 W) architectures. Because mobile traffic drops to 15–20% of peak capacity at night, AI-RAN clusters leave substantial GPU compute idle during off-peak hours. We show that time-sharing these idle GPUs for external Large Language Model (LLM) inference (the AI-on-RAN model) generates up to an 8.2x return multiple on marginal hardware costs when clusters are sized for 10-year peak demand. The model also evaluates infrastructure sharing and energy control. Multi-Operator Core Network (MOCN) active sharing and Neutral Host deployments cut the rural break-even ARPU threshold by more than 50% (from €35–40 down to €12–15 per month, supporting UN SDG 9.1). Meanwhile, RAN Intelligent Controller (RIC) sleep policies reduce carbon intensity per unit revenue by 55–60% over seven years, and direct GPU capacity leasing lets cloud hyperscalers offset 40–50% of BBU costs without mandatory telecom levies.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24422605
Primary Topic
Software-Defined Networks and 5G
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Techno-Economic Assessment and Return on Investment of Open RAN and AI-RAN Architectures in 6G Deployments

Ivan Gomanchenko
Zenodo (CERN European Organization for Nuclear Research)
Software-Defined Networks and 5G
article

Techno-Economic Assessment and Return on Investment of Open RAN and AI-RAN Architectures in 6G Deployments

Ivan Gomanchenko
article en

Abstract

High CapEx and flat ARPU in 5G networks create serious financial risks for 6G rollouts, where the Radio Access Network (RAN) still accounts for 60% to 70% of Total Cost of Ownership (TCO). This paper models a 10-year TCO and Return on Investment (ROI) comparison between COTS-based Open RAN (Intel FlexRAN VRB1, 300 W) and GPU-based AI-RAN (NVIDIA GH200 Grace Hopper, 1,200 W) architectures. Because mobile traffic drops to 15–20% of peak capacity at night, AI-RAN clusters leave substantial GPU compute idle during off-peak hours. We show that time-sharing these idle GPUs for external Large Language Model (LLM) inference (the AI-on-RAN model) generates up to an 8.2x return multiple on marginal hardware costs when clusters are sized for 10-year peak demand. The model also evaluates infrastructure sharing and energy control. Multi-Operator Core Network (MOCN) active sharing and Neutral Host deployments cut the rural break-even ARPU threshold by more than 50% (from €35–40 down to €12–15 per month, supporting UN SDG 9.1). Meanwhile, RAN Intelligent Controller (RIC) sleep policies reduce carbon intensity per unit revenue by 55–60% over seven years, and direct GPU capacity leasing lets cloud hyperscalers offset 40–50% of BBU costs without mandatory telecom levies.

Zenodo (CERN European Organization for Nuclear Research)
Saint-Petersburg State University of Telecommunications (RU)
Software-Defined Networks and 5G
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon