Telecom operators are accelerating plans for autonomous networks that can understand intent, reason over tradeoffs and act without manual intervention, with network automation identified as the top artificial intelligence use case for investment and return on investment in Nvidia’s latest State of AI in Telecommunications report. Moving from basic automation to full autonomy requires reasoning models and artificial intelligence agents fine tuned on telecom data, as well as an end to end agentic system in which telco network models and agents communicate with each other and use simulation tools to validate actions. Ahead of Mobile World Congress Barcelona, Nvidia introduced an open Nvidia Nemotron based large telco model, a detailed implementation guide for building reasoning agents, and new Nvidia Blueprints targeting energy saving and network configuration using multi agent orchestration, all of which are being released as open resources through GSMA’s Open Telco AI initiative.
The new open source, 30-billion-parameter Nvidia Nemotron large telco model, developed with AdaptKey Artificial Intelligence and based on the Nemotron 3 family, is tuned on open telecom datasets including industry standards and synthetic logs so it can understand telecom specific terminology and reason through workflows like fault isolation, remediation planning and change validation. As an open model, it provides operators with transparency into its training and data, supports secure on premises deployment, and allows telcos to extend telecom tuned reasoning with their own network and operational data while maintaining control over security and data governance. Nvidia and Tech Mahindra have also published an open source guide that explains how to fine tune domain specific reasoning models and construct agents capable of safely executing network operations center workflows by converting expert resolutions into step by step procedures and structured reasoning traces that capture each action, tool invocation, outcome and decision.
To operationalize autonomous behavior, Nvidia is expanding its blueprint portfolio with a focus on closed loop operation, where models interpret the network, agents act on operator intent, and simulation feeds results back to refine decisions. The Nvidia Blueprint for intent-driven RAN energy efficiency combines these elements to help operators reduce power usage in 5G radio access networks while preserving quality of service, integrating Viavi’s TeraVM Artificial Intelligence RAN Scenario Generator to produce synthetic data on cell utilization, user throughput and traffic patterns that an energy planning agent uses to propose policies, which are then validated in simulation without touching live configurations. The Nvidia Blueprint for telco network configuration is already in production with operators: Cassava Technologies is building the Cassava Autonomous Network with three cooperating agents for monitoring, change execution and impact assessment across Africa’s multi vendor environment, while NTT Data is using the approach with a tier 1 operator in Japan so an artificial intelligence agent can regulate traffic by admitting users based on real time demand and adapting decisions as conditions stabilize. Nvidia and BubbleRAN are now enhancing the configuration blueprint with the Nvidia NeMo Agent Toolkit and BubbleRAN Agentic Toolkit to orchestrate multiple agents across the RAN, with BubbleRAN integrating these into its Opti-Sphere platform so monitoring, configuration and validation agents can run flexibly across containers, connect to metric and traffic tools, and continuously propose and validate configuration changes, an approach that Telenor Group will first deploy to strengthen its 5G network for Telenor Maritime.
