T-Mobile has completed an AI-RAN voice call on its live commercial 5G network alongside Nokia and Nvidia, marking a significant step toward integrating artificial intelligence directly into mobile network infrastructure. The companies describe the Bellevue, Washington, demonstration as the world’s first AI-RAN voice call, moving the technology beyond laboratory testing into an operating telecommunications environment. The milestone brings together Nokia’s radio access technology, Nvidia’s accelerated computing platform and T-Mobile’s commercial network, offering an early indication of how carriers could redesign network architecture for increasingly demanding AI workloads. For the telecommunications industry, the test signals a shift in the AI-RAN conversation from long-term architectural ambitions toward practical deployment and performance validation.

T-Mobile Moves AI-RAN From Laboratory to Live Network

The call originated from a commercial cell site in Bellevue, where T-Mobile is headquartered, and traveled across the operator’s live production network. The setup combined Nokia Massive MIMO radio equipment, Nokia 5G anyRAN software and server infrastructure powered by Nvidia’s Aerial RAN Computer platform. Nokia confirmed in its announcement that the site connected to a commercial 5G core and commercial user equipment, placing the demonstration within an operational network environment rather than an isolated laboratory configuration. That distinction matters because operators must establish that AI-enabled radio functions can coexist with existing network operations before they can justify broader infrastructure investments.

AI-RAN, or artificial intelligence radio access network technology, seeks to integrate AI capabilities into the infrastructure responsible for connecting mobile devices to telecommunications networks. The approach could allow operators to optimize radio performance, improve spectrum utilization and support AI inference workloads on shared computing infrastructure. However, a successful voice call represents an initial technical milestone rather than proof that AI-RAN can deliver its proposed efficiency gains across an entire commercial network. Operators will still need to validate performance, reliability, energy consumption and operating costs under different traffic conditions before committing to large-scale deployments.

Nokia Targets Greater Spectrum Efficiency Ahead of 6G

Nokia views AI-RAN as a central component of its strategy to transform radio access networks into software-defined, AI-native platforms. Such an architecture could give operators greater flexibility to introduce new capabilities through software and computing infrastructure rather than relying exclusively on dedicated radio network equipment. The strategy also aligns with the industry’s preparations for 6G, although the commercial rollout of AI-RAN products remains a separate milestone from the development of next-generation mobile standards. For infrastructure planners, the attraction lies in the possibility of using existing spectrum and network assets more efficiently while creating a foundation for future AI-enabled services.

“AI-RAN represents an important evolution of the radio network. Nokia’s vision is to evolve the RAN into an increasingly software-defined, AI-native platform, creating a common foundation for future AI-enabled services and the evolution toward 6G,” said Pallavi Mahajan, chief technology and AI officer, Nokia. “With AI-RAN, we are on a path to 2x spectral efficiency, an ambition that puts more capacity from the same spectrum at the heart of what comes next.”

Field Trials Will Test AI Workloads Alongside RAN Functions

The companies plan to expand their field trials to cover additional AI-for-RAN algorithms, spectrum bands and AI inference workloads running alongside radio access network functions on a shared accelerated computing platform. The next stage will test how the architecture behaves under more varied operating conditions and whether it can support additional use cases without undermining core network functions. Those trials will help determine how the technology performs beyond a single voice-call demonstration and what engineering changes operators may need before deployment at scale. They will also provide a clearer basis for assessing the relationship between computing utilization, network performance and the economics of AI-enabled radio infrastructure.

“The next phase of field trials will build on this milestone, exploring additional spectrum bands and the ability to support AI inferencing workloads alongside RAN functions. Each step brings us closer to understanding how AI-RAN can deliver meaningful benefits in real-world network environments and ultimately operate at scale,” added Pallavi.

T-Mobile’s Bellevue Innovation Center Supports the Strategy

T-Mobile has been developing its AI-RAN capabilities through its AI-RAN Innovation Center in Bellevue, announced in September 2024. The center reflects the operator’s interest in exploring how AI could improve network performance and support new services as telecommunications infrastructure evolves. At Mobile World Congress, T-Mobile chief technology officer John Saw indicated that the company hoped to have trial units available to test during the year, as Nokia outlined its AI-RAN roadmap with Nvidia and several carrier partners. The completed call now provides a concrete demonstration of progress toward those earlier testing ambitions.

The companies are also pursuing opportunities that connect AI infrastructure with applications at the network edge. Earlier in 2026, T-Mobile, Nvidia and Nokia announced work with additional partners to deliver physical AI applications over distributed Edge AI networks. Physical AI encompasses applications that interact with the real world, including systems that rely on connected devices, sensors and machines to interpret information and respond to changing conditions. Integrating these applications with distributed computing infrastructure could create new opportunities for telecom operators, although the commercial prospects will depend on application demand, latency requirements and the availability of suitable edge resources.

Nokia does not expect its first commercial AI-RAN products to go live until 2027, leaving further development and validation between the Bellevue demonstration and broader deployment. That timeline makes the current trial an early indicator of technical progress rather than evidence of an established commercial market. Before operators can introduce AI-RAN at scale, they will need to assess equipment costs, software maturity, network reliability, power requirements and compatibility with existing infrastructure. The results will help determine whether the technology can move from experimental deployments into mainstream network planning.