Northeastern and SoftBank Demo Agentic AI-RAN
Northeastern University, SoftBank, and partners demonstrated an intent-driven, AI-native Radio Access Network (RAN) at MWC Barcelona 2026. The system is powered by a Large Telecom Model (LTM) and uses autonomous agentic AI to manage network functions, representing a new approach to telecommunications infrastructure.
- The underlying Large Telecom Model (LTM) from SoftBank is a generative AI foundation trained on vast datasets, including network operational know-how, to automate tasks like base station configuration, reducing completion times from days to minutes with over 90% accuracy in trials. - This AI-RAN initiative is a key component of the AI-RAN Alliance, co-founded by SoftBank and Northeastern University, which aims to integrate AI and RAN processes to create new AI-driven revenue opportunities and deploy AI services at the network edge. - Ztouch Networks, a Northeastern University spinoff, provides the orchestration platform that enables the deployment and management of AI and RAN workloads on shared GPU resources, a concept referred to as AI-and-RAN. - Keysight Technologies provides the critical testing and validation framework, using high-fidelity emulation and digital twins to ensure that the AI and RAN workloads can coexist and perform reliably under realistic network conditions before deployment. - A primary business driver for this architecture is the monetization of underutilized network capacity; NVIDIA and SoftBank estimate that for every $1 of CAPEX invested in AI-RAN infrastructure, operators could earn roughly $5 in AI inference revenue over five years. - The "agentic" aspect of the AI refers to autonomous agents that can understand intent, reason across complex systems, and execute multi-step workflows without direct human intervention, moving beyond simple automation to dynamic, goal-oriented decision-making. - This approach is part of a broader industry trend toward AI-native 6G networks, where AI is not an add-on but a foundational component for optimizing everything from spectral efficiency to power consumption and enabling new applications like integrated sensing and communication (ISAC). - For platform engineering leaders, the architecture introduces new challenges and opportunities in API design and orchestration, requiring northbound APIs for developers to submit workloads and express intent, and a sophisticated scheduler to manage resource allocation between time-sensitive RAN functions and AI tasks.
