6G and AI-Native Networks Tested for Remote Compute
The architecture for future autonomous systems is taking shape as DOCOMO and NTT successfully demonstrated low-latency AI video analytics using remote GPUs over an advanced network. This vision of offloading heavy AI tasks is complemented by Nokia's push to build AI-native 6G networks from the ground up, signaling a future of distributed, network-aware compute for aerospace platforms.
The DOCOMO and NTT trial successfully integrated remote GPUs over a commercial 5G Standalone network and NTT's IOWN (Innovative Optical and Wireless Network) All-Photonics Network. This setup enabled AI inference processing to be controlled directly from the network, a key step in centralizing control of both communication and computation for future services. The demonstration achieved a combined communication and AI analysis latency that meets the safety parameters for remote robot control near humans, as defined by ISO/TS 15066. The architecture split the AI inference into pre-processing and execution stages, transmitting the pre-processed data to distant GPUs. Priority control functions within DOCOMO's commercial 5G core, running on AWS, were crucial for managing the low-latency transmission. This in-network computing approach is designed to overcome the limitations of traditional distributed AI, where performance is highly dependent on the physical proximity of GPU resources. Nokia's AI-RAN (Artificial Intelligence Radio Access Network) strategy, developed in partnership with NVIDIA, aims to integrate AI capabilities directly into the radio access network infrastructure. This approach uses general-purpose hardware to run both cellular and AI workloads concurrently, turning the RAN from a single-purpose system into a multi-purpose cloud infrastructure. The goal is to enhance spectral efficiency, manage network traffic dynamically, and enable new AI-based services at the network edge. At Mobile World Congress 2026, Nokia showcased these AI-RAN capabilities with partners like Dell, Red Hat, and SuperMicro. The demonstrations included using spare GPU capacity in the distributed network for external AI compute services, effectively monetizing underutilized resources. This aligns with the broader industry vision for 6G to not just provide faster speeds but to also serve as a distributed sensing and computation platform for autonomous systems. This shift towards AI-native networks involves embedding AI/ML models directly into the radio signal processing layer to optimize performance in real-time. It represents a move away from specialized, purpose-built hardware towards more flexible, software-defined platforms. The AI-RAN Alliance, an industry consortium, is working to establish common frameworks and architectures to accelerate this transition. For aerospace, 6G and AI-native networks promise to enhance situational awareness through real-time data from multiple sensors and to provide the ultra-reliable connectivity needed for autonomous drones and vehicles. The technology is expected to support the Internet of Battlefield Things (IoBT) and enable advanced predictive analytics and more resilient cybersecurity measures. This evolution will be critical for enabling complex, time-sensitive operations in contested environments with minimal human intervention.
