Telcos see near-term AI-RAN value but caution against speculative build-out
Operators including Optus, Orange, and SK Telecom are focusing on AI-driven improvements to existing radio networks rather than speculative AI-native architectures. ZTE and Samsung also weigh in, with ZTE saying the business case for RAN-for-AI is immature. The industry warns against repeating the 5G pattern of building capacity before applications exist.
The Intelligent RAN Forum this week surfaced a clear split in operator thinking on AI-RAN. Optus reported tangible spectral efficiency gains of up to 25% on its Ericsson infrastructure, while Orange has moved AI-RAN into live network testing. Both frame AI as a tool to optimize existing radio assets rather than a foundation for new architectures.
ZTE cautioned that the business case for dedicating radio networks to AI workloads remains unproven, with demand confined to narrow use cases. SK Telecom warned carriers against repeating the 5G-era mistake of building capacity before applications materialize, citing MEC and URLLC as cautionary examples. Samsung continues exploring longer-term AI-native concepts with Korean operators.
The industry's shift toward pragmatic AI-for-network improvements could deliver more efficient spectrum use and lower operational costs, potentially translating into better coverage or pricing for consumers. However, if carriers overinvest in speculative AI-native infrastructure, the costs could mirror 5G's build-first pattern, where consumers bore the expense of capacity that took years to fill. The outcome may shape how quickly AI-enhanced connectivity reaches everyday users.