On October 21, NVIDIA disclosed that T-Mobile had completed the world’s first commercial-network AI-RAN call, with Deutsche Telekom, BT, and Elisa also launching AI-RAN trials — major global operators are bringing accelerated computing into telecom networks. On the same day, Shenzhen Tongyu Communication and Jiaxian Communication announced a strategic partnership, with Tongyu planning to acquire approximately 25% of Jiaxian for around RMB 300 million. Jiaxian simultaneously released an AI-compute distributed base station, a 6G integrated AI-RAN base station supporting local AI agent deployment, and a 6G edge-terminal AI agent combining 6GHz radio access with edge AI compute.
AI-RAN is not simply about putting GPUs into base stations. Its key lies in task layering: model training and heavy inference stay in the data center, while latency-sensitive, data-intensive control and inference tasks requiring on-site closed loops sink down to base stations, campuses, or edge equipment rooms. Take robotic grasping as an example — if judgments about material, weight, and posture all round-trip to the cloud, real-time control is hard to guarantee; with edge nodes completing inference nearby, the robot body can reduce its compute, power, and cost burden. The Omdia forecast cited by NVIDIA and Nokia puts the cumulative AI-RAN market opportunity above USD 200 billion by 2030. Jiaxian Communication plans to advance experimental and production-unit development in 2027, large-scale trials in 2028–2029, and a development peak around 2030.
From an industry landscape perspective, AI-RAN is restructuring the radio access network value chain: base station BOM structures, power and thermal design, software stacks, and operations models will all change, and a collaborative system of chip platform plus equipment plus network software plus operator scenarios is gradually replacing the pure equipment-vendor-dominated model. For operators, sites, spectrum, backhaul networks, and enterprise customer resources are being repriced in the edge computing era, shifting business models from traffic pipes to connectivity plus compute. This is an architectural rehearsal ahead of 6G standardization — how interfaces open up, how compute is orchestrated, and how AI tasks are distributed across cloud, edge, and device will determine the telecom industry division of labor for the next decade.
Junyi Microelectronics has long focused on edge computing hardware. Its 5G-EDGE-3399Pro and AI-BOX-3588 edge computing hosts are built for exactly these compute-sinks-to-the-site scenarios, handling local AI inference and real-time control in campuses, production lines, and mobile carriers. Together with J-3568/J-3576/J-3588 core boards and K-series edge intelligent motherboards, Junyi provides industry customers with a one-stop hardware foundation spanning sensing, connectivity, and edge inference. The base-station-as-compute-gateway trend represented by AI-RAN strongly resonates with the company’s product layout in 5G-edge convergence.
Source: Communications World: NVIDIA’s AI-RAN Offensive — Base Stations Become Computing Gateways (2026-10-24)