NTT DoCoMo launched Japan’s first commercial 5G core network on Amazon Web Services (AWS), while revealing a new in-network computing capability in its mainline 5G core service (5GC).
The Japanese operator launched the AWS instance of its 5GC during the Mobile World Congress (MWC) 2026 event. NEC supplied the cloud-native software stack for the service, with the firm redesigning DoCoMo’s overall core architecture so it could run in hybrid form.
A construction and operations framework based on infrastructure-as-code (IaC) and continuous integration/continuous delivery (CI/CD) was implemented for AWS, integrating AWS managed services such as CloudFormation, CodeBuild, and CodePipeline into the design. Tests were run on AWS and a DoCoMo proprietary virtualization platform.
Keisuke Suzuki, director of operation support system at NTT DoCoMo, stressed to SDxCentral that the AWS relationship goes beyond hosting into AI governance, data analysis, and virtualized radio access network (vRAN) integration.
“We’ve been working on 5G core projects on AWS for a couple of years,” Suzuki explained. “We leverage Amazon Bedrock agents, and they offer a number of governance tools for us.”
Suzuki expounded on DoCoMo’s agentic framework, citing a network recovery agent that pulls data from more than one million devices on its network.
“I think this is one of the largest data sets for network service assurance," Suzuki said. "So AI agents can handle a huge amount of data and easily analyze it, and identify the root cause and create recovery actions based on its discoveries.”
The director also revealed DoCoMo uses AWS for vRAN support, highlighting an MWC demo of edge computing running AI applications alongside radio workloads.
“We can demonstrate that in the same sidebar, the builder application is running, and also the application is running," Suzuki said. "As such, workloads running on AWS can be easily deployed into our vRAN.”
DoCoMo on the edge
Following the launch of the AWS instance for Japan, DoCoMo and parent company NTT revealed an in-network computing capability within the mainline 5GC instance dubbed "Inc Edge."
According to an MWC-themed blog from NTT Group, the functionality connects the mobile network to NTT's all-photonics IOWN APN service to help coordinate AI inference processing in tandem with traffic control.
Named after the NTT-led Innovative Optical and Wireless Network (IOWN) Global Forum, IOWN APN aims to address growing capacity and rising power consumption. With Inc Edge able to determine how and where inference tasks are executed, data is routed via the APN to remote GPU services where inference takes place.
At MWC, this was shown with video captured by a device before being transmitted over a commercial 5G network. Priority control within the 5G core was applied to AI-related traffic, while the APN ensured high-bandwidth and low-latency connectivity. Pre-processing was performed locally before forwarding data, ensuring that only necessary intermediate results were sent across the APN to remote GPUs.
This approach, NTT said, minimizes unnecessary data transfers and enhances the efficiency of geographically distributed compute resources, while giving operators like DoCoMo greater flexibility in GPU placement and capacity management for large volumes of video or sensor data that require rapid analysis.
Such data is likely to rise with the tide that is physical AI, referring to AI-native, sensor-laden technology such as robots and drones. In time for the MWC event, NTT subsidiary NTT Data partnered with Ericsson to galvanize private 5G enterprise adoption and spur on advanced physical AI and edge-based AI use cases.
Paul Bloudoff, senior director for edge AI and 5G strategic client enablement at NTT Data, said both firms believe that "AI is living at the edge," processing sensor data from cameras, vibration, and even sound generating data from radar, light detection and ranging (lidar), and other sensing sources.
"Wi-Fi is unreliable," Bloudoff stated. "Ericsson is one of the top names in private 5G, everyone knows the story at this point. Ericsson has a wonderful portfolio of global RAN available to us. They have a very good core, just like Nokia does, but there are reasons to go with one or another, dependent on use case.
"Sometimes some cores scale better with different types of radios," Bloudoff continued. "The radio inventory is different from vendor to vendor, and that's why clients work with people like NTT to help to come up with the best solution for their city, their airport, their port, their logistics center, their highways, you name it."
Bloudoff believes AI is driving private 5G uptake, citing the customer case study of chemical manufacturers Celanese.
"We have AGVs (automated guiding vehicles) moving around their refinery plants to start to digitize what had traditionally done by humans as they walk around and note down temperature and pressure gages," Bloudoff explained. "You can now send spot a robotic dog to do that on a reliable basis. At the same time it's doing that, it can monitor for security, intrusions, for anything that's out of place. So the use cases for private 5G grow over time as well."
