Nokia Transforms South Korean Central Offices into AI Edge Data Centers to Maximize Network Investment Efficiency via Unified AI RAN Platforms
South Korea will be the first market where Nokia will deploy Artificial Intelligence RAN. This was announced by Nokia at the Amplify press conference in Seoul, as reported by ZDNet Korea. Dr. Cho Bong yeol of Nokia Korea pointed out that the highly developed centralized office architecture and fast technological uptake in the region are two main factors behind this move. The execution of this plan involves using existing telecom structures for local processing of AI workloads and generating a new income source for network operators.
The core concept involves AI RAN, where AI computing is incorporated directly into the wireless access network itself. Such an architecture provides the ability for local telecom offices to carry out workload intensive processes, such as inference, before the data even reaches the cloud data center. Nokia, through its strategic partnership with Nvidia, has a 1 billion USD chip investment and is studying ways in which GPU computing can be combined with the network itself.
As advanced software models like Agentic AI and Physical AI become more common, the demand for real time computational inference will increase. Han Hyo chan, the Chief Technology Officer of Nokia Korea, explained that these calculations must be processed as close to the user as possible to maintain low latency. By installing GPUs into the digital unit servers located within existing central stations, South Korean telecom operators can essentially transform their local facilities into small or medium sized AI data centers.
This edge computing model enables local telecom companies to build new business models with large cloud providers abroad. Today, the three largest domestic carriers SK Telecom, KT, and LG Uplus are working directly with Nokia to develop capabilities for AI RAN and 6G networks. Working with Nokia on this effort is also a global group of partners including T Mobile in the US, SoftBank in Japan, and Indosat in Southeast Asia, to demonstrate the commercial applicability of the technology.
Compared to traditional network deployment, telcos typically need to incur Capex investments on both regular wireless access networks and AI GPU clusters. With AI RAN, telcos have overcome this bottleneck by fusing communication and AI computing on one unified platform, which enables more flexible and dynamic resource deployment according to real time demand.
When the network experiences any normal demand, the available onboard computing power is used to support 5G or 6G cellular capacity. When demand for digital services peaks, that same available computing power can be increased immediately to support AI inference and token transactions. This capability optimizes the investment return on network infrastructure by enabling telecom companies to receive direct revenue from their physical hardware investments.
