Telcos are spending billions to manage increasingly complex networks. NVIDIA thinks AI agents — powered by LLMs — can bring that cost down.
Telecom networks are under constant pressure: growing user demands, massive data volumes, and the relentless need for higher performance. But managing those networks is expensive — nearly $1.3 trillion in global spending last year alone, with $295 billion in capital expenditures and over a trillion in operating costs.
A major part of this cost comes from manual, rule-based network configuration tasks: tuning parameters, optimizing traffic flow, balancing loads, and mitigating interference — often in real time. These tasks are traditionally handled by engineers using rigid logic and static policies.
But NVIDIA wants to change that.
At GTC Paris, the company unveiled its latest AI Blueprint — a free, open-source framework designed to automate network configuration for telcos using large language models (LLMs) and agentic AI.
This could be a turning point for telcos on the road to fully autonomous networks.
NVIDIA’s new AI Blueprint for Telco Network Configuration brings the power of customized LLMs directly into the network layer.
Instead of relying on predefined rules, telcos can now use AI agents that:
- Understand dynamic network data
- Make intelligent, real-time decisions
- Continuously optimize network settings
- Balance trade-offs like latency vs. bandwidth or energy use vs. performance
Built using NVIDIA’s NIM microservices and trained on 5G data from BubbleRAN, the AI agents can autonomously configure and update key parameters that affect quality of service, reliability, and energy consumption — all without human input.
NVIDIA AI Blueprints, available on build.nvidia.com, provide developers with:
- 📄 Reference code
- 🛠️ Deployment tools
- 📚 Technical documentation
- 🎯 End-to-end architecture for building agentic AI solutions
The Telco Network Configuration Blueprint in particular is trained on domain-specific datasets, including BubbleRAN 5G O-RAN platform data, and enables LLMs to understand how to:
- Set optimal signal-to-noise ratios
- Maintain consistent bitrates
- Adapt configurations as traffic patterns and mobility shift throughout the day
This continuous learning loop allows AI agents to adaptively tune network performance — a key requirement in the era of 5G and future 6G networks.
Norway-based Telenor Group, with over 200 million customers worldwide, is the first telco to pilot this blueprint.
“The blueprint is helping us address configuration challenges and enhance quality of service during network installation,” said Knut Fjellheim, CTO at Telenor Maritime. “Implementing it is part of our push toward network automation and follows the successful deployment of agentic AI for real-time network slicing.”
Telenor’s use case — automating private 5G maritime networks — showcases the blueprint’s real-world potential in both performance optimization and operational cost reduction.
✅ For Telcos:
- Slash manual workloads and costs
- Improve quality of service (QoS)
- Speed up deployment of new 5G/6G services
- Reduce human error and improve resilience
✅ For Developers:
- Build, customize, and deploy telco-specific LLMs
- Leverage NVIDIA NIM microservices for deployment
- Tap into pre-trained models and domain data
✅ For the Industry:
This marks a key shift from static network policies to dynamic, AI-driven automation — a move that could fundamentally reshape how telecom infrastructure is managed.
As networks grow more complex, telcos can’t afford to rely on manual processes. With LLMs now capable of understanding domain-specific configurations and adapting in real time, agentic AI offers a blueprint for a smarter, more cost-efficient future.
And NVIDIA is offering that blueprint for free — right at the intersection of AI innovation and real-world telecom needs.