AI-RAN is showing early financial value inside mobile networks. Yet the clearest returns come from operations, not new services.
That was the main view from an Intelligent RAN Forum panel. Speakers included Güneş Kesik of Turkcell, Gabriel Ionita of Deutsche Telekom, and Rob Hughes of 1Finity. The session was moderated by Tom Camp.
The panel agreed on today’s strongest use cases. Operators can reduce energy use. They can find network faults faster. They can also improve planning and capacity use.
“We see the clearest value in energy optimization, better resource use, and operational automation,” Kesik said.
This matters because radio networks are expensive to run. Energy use remains a major operating cost. AI tools can adjust resources based on real traffic needs. That can cut waste without hurting service quality.
Fault handling also offers a strong case. Large networks often generate alarm floods during failures. AI can help find the real cause faster. Hughes said this can reduce repair analysis from hours to minutes.
However, the revenue story remains less clear. Operators want AI-RAN to support new edge services. These services could help factories, vehicles, and enterprises. Yet businesses must first show real buying interest.
Ionita focused on practical value from existing assets. “Making better use of the network we already have,” he said.
That view reflects a wider market reality. Many carriers face flat radio access network spending. They must extract more value from existing infrastructure. AI helps when it delays new capital spending.
Kesik urged operators to measure AI at network level. Energy savings alone may not tell the full story. Better capacity use can also delay new site investments.
Still, the panel warned about hidden challenges. None of the speakers named hardware as the biggest issue. Instead, they highlighted skills, older systems, and lifecycle work.
AI-RAN needs radio expertise and IT knowledge. It also needs teams who understand data centers. That blend remains hard to find in many operators.
Legacy system integration creates another burden. “The things that get forgotten are the integrating with the legacy systems,” Ionita said.
Kesik described the balance clearly. “The complexity we remove through automation should be greater than the complexity we introduce with AI,” Kesik said.
Open RAN experience may help here. Multi-vendor work has already taught operators useful lessons. Standards and automation can reduce future integration effort.
The bigger strategic question remains unresolved. Developers want edge infrastructure before building applications. Operators want applications before funding widespread edge builds.
Hughes argued for customer-led deployments. Operators should build where an enterprise is ready to pay. That lowers risk and tests the business model early.
GPU services may become another opportunity. But Ionita framed them as a partnership area. Operators do not need to replace cloud giants.
For now, AI-RAN’s value sits in better network operations. New revenue may follow. But it still needs clear demand.
