Analyst weighs when AI-RAN upgrades deliver operator value
An analyst piece examines when AI-RAN investments pay off for operators, noting that gains depend on where cells sit on their capacity curves. SoftBank and Ericsson reported roughly 10% improvements in spectral efficiency and downlink throughput from AI link adaptation on a commercial 5G network, using existing baseband hardware. Nokia is also reporting gains and pursuing accelerated computing, but the business case for new hardware depends on utilization and inference costs.
An analyst notes AI-RAN is producing measurable radio improvements. SoftBank and Ericsson found roughly 10% average gains in spectral efficiency and downlink throughput on a live commercial 5G network, using baseband equipment already deployed. Nokia reports over 20% combined spectral-efficiency gains from AI-assisted algorithms on current platforms, spanning link adaptation, channel estimation and carrier aggregation.
Nokia is also shifting these functions to accelerated merchant silicon, including NVIDIA Aerial, with roadmap targets of 50% by 2027 and over 100% by 2028. Those figures are planned, not commercial results. The analyst stresses that value depends on cell loading, subscription pricing and scope, and inference costs.
Operators could use AI-RAN gains to postpone some capacity upgrades, potentially lowering network costs or redirecting spending. Mobile users in congested areas may see improved throughput or fewer slowdowns, while lightly loaded regions may notice little change. Equipment vendors and software providers may benefit if subscription models scale, though accelerated platforms could raise dependence on inference costs and specialized hardware. The broader social effect may depend on whether these efficiency gains translate into wider coverage, better service, or simply deferred investment.