NVIDIA Advances Autonomous Networks With Agentic AI Blueprints and Telco Reasoning Models
Source: NVIDIA AI Blog
Why Autonomy Matters
-
Automation ≠ Autonomy
- Automation executes predefined workflows.
- Autonomy requires the network to understand operator intent, reason over trade‑offs, and decide on actions.
-
Key Enablers
- Reasoning models and AI agents fine‑tuned on telecom data.
- An end‑to‑end agentic system that includes:
- Telco network models.
- AI agents that communicate with each other.
- Network simulation tools for action validation.
Recent NVIDIA Announcements (Ahead of MWC Barcelona)
| Announcement | What It Is | Why It Matters |
|---|---|---|
| Open NVIDIA Nemotron‑based Large Telco Model (LTM) | First open‑source, large‑scale telco model built on the Nemotron architecture. | Provides a foundation for building reasoning agents that understand network dynamics. |
| Comprehensive Guide for Building Reasoning Agents | Step‑by‑step implementation guide. | Accelerates development of autonomous‑network agents. |
| NVIDIA Blueprints | Pre‑designed, multi‑agent orchestration workflows for: • Energy‑saving • Network configuration | Gives operators ready‑to‑use, proven patterns to move toward autonomy. |
Collaboration with GSMA
- Open Telco AI Initiative (launching tomorrow) – a joint effort between NVIDIA and the GSMA, the global mobile communications association.
- NVIDIA will release, through GSMA, the following open resources:
- The open‑source LTM.
- The implementation guide for reasoning agents.
- The agentic AI blueprints for energy efficiency and network configuration.
These resources aim to help operators accelerate the transition to autonomous networks and unlock new levels of efficiency, reliability, and innovation.
Open Nemotron 3 Large Telco Model Brings Reasoning to Telecom
For telcos to successfully operationalize generative and agentic AI across their operations, AI models must be able to understand telecom terminology and reason through complex workflows. NVIDIA has partnered with AdaptKey AI to release a new open‑source, 30‑billion‑parameter NVIDIA Nemotron LTM that operators worldwide can use to build autonomous networks.
Key Features
- Built on the NVIDIA Nemotron 3 family of foundation models.
- Fine‑tuned by AdaptKey AI using open telecom datasets, including industry standards and synthetic logs.
- Optimized for:
- Fault isolation
- Remediation planning
- Change validation
Benefits for Telcos
- Full transparency – Access to training data and methodology.
- Secure, on‑premises deployment – Run agents directly within the network.
- Customizable reasoning – Extend and adapt the model with proprietary network and operational data.
- Path to autonomous operations without compromising data control or security.
Teaching AI Agents to Reason Like Network Engineers
NVIDIA and Tech Mahindra have released an open‑source guide that shows telecom operators how to fine‑tune domain‑specific reasoning models and build agents capable of safely executing Network Operations Center (NOC) workflows.
Key Takeaways
- Focus on high‑impact, high‑frequency incidents – Identify the incident categories that occur most often and have the greatest operational impact.
- Translate expert resolutions into step‑by‑step procedures – Break down each resolution into discrete actions, tool calls, outcomes, and decisions.
- Create structured reasoning traces – Capture the procedure as a “thinking example” that records every step the engineer would take, why it is performed, and what result is expected.
These reasoning traces become the training data that teach the model not only what to do, but why a particular sequence of checks and fixes is safe and effective.
How It Works
- Collect expert knowledge – Gather incident logs, run‑books, and engineer notes.
- Convert to reasoning traces – Encode each resolution as a structured trace (e.g., JSON or YAML) that includes:
- Action description
- Tool or API invoked
- Expected outcome
- Decision logic for the next step
- Fine‑tune with NVIDIA NeMo‑Skills – Use the NeMo‑Skills pipeline to train a reasoning model on the traces.
- Deploy as a telco‑specialized AI agent – The resulting model can reason through new incidents, suggest actions, and execute safe NOC workflows autonomously.
Benefits
- Accelerated incident resolution – AI agents can handle routine, repetitive tasks, freeing engineers for complex problems.
- Consistency and safety – Structured reasoning ensures actions follow vetted, repeatable procedures.
- Scalable expertise – Knowledge from senior engineers is codified and shared across the organization.
For a deeper dive, refer to the full NVIDIA‑Tech Mahindra guide linked above.
Maximizing Energy Efficiency with the New Intent‑Driven Energy‑Saving Blueprint
Autonomous networks rely on a closed‑loop workflow:
- Model – Understand the network state.
- Agent – Act on high‑level intent.
- Simulation – Feed results back to validate and refine decisions.
The new NVIDIA Blueprint for intent‑driven RAN energy efficiency ties these three pieces together, enabling operators to systematically cut power consumption in 5G Radio Access Networks (RAN) while preserving Quality of Service (QoS).
Key Components
| Component | Role | Link |
|---|---|---|
| VIAVI TeraVM AI RAN Scenario Generator (AI RSG) | Generates synthetic network data (cell utilization, user throughput, traffic patterns, etc.) and exports it in a simple, queryable format. | VIAVI blog – AI RSG |
| Energy‑Planning Agent | Analyzes the synthetic data, creates energy‑saving policies, and feeds them back into AI RSG for simulation. | — |
| Closed‑Loop Validation | Operators can safely test policies in simulation before applying any changes to live networks, ensuring intent is met without impacting subscribers. | — |
How It Works
- Data Generation – AI RSG produces realistic, synthetic RAN scenarios covering a wide range of traffic conditions.
- Policy Reasoning – The energy‑planning agent evaluates the data, identifies opportunities (e.g., dynamic cell sleep, power‑level scaling), and proposes policies.
- Simulation & Validation – Proposed policies are run through AI RSG to predict their impact on power usage and QoS.
- Iterative Refinement – Results feed back to the agent, which adjusts policies until the desired intent (energy reduction with QoS guarantees) is achieved.
Benefits
- Predictable Energy Savings – Quantifiable reductions before any live deployment.
- Zero Service Disruption – All testing occurs in a sandboxed environment.
- Scalable Intent‑Driven Automation – Policies can be rolled out across thousands of sites with minimal manual effort.
- Continuous Improvement – Closed‑loop feedback enables ongoing optimization as traffic patterns evolve.
Ready to start reducing your RAN power footprint? Explore the full blueprint and begin building intent‑driven energy‑saving workflows today.
Telcos Put the NVIDIA Blueprint for Network Configuration to Work
The NVIDIA Blueprint for telco network configuration is being adopted by operators worldwide.
Notable Deployments
| Operator / Partner | How the Blueprint Is Used | Key Benefits |
|---|---|---|
| Cassava Technologies | • Built the Cassava Autonomous Network, an agentic platform for Africa’s multi‑vendor mobile environment. • Implements three agents: 1️⃣ Monitor network & recommend config changes. 2️⃣ Apply changes with documentation & governance. 3️⃣ Assess impact & safely roll back unintended effects. | • Optimizes heterogeneous networks. • Enables automated, auditable configuration cycles. |
| NTT DATA | • Integrates the blueprint to add intelligence to traffic regulation. • Helps networks manage surges when users reconnect after outages. • Deployed with a tier‑1 operator in Japan. | • Improves resilience during demand spikes. • Turns manual tuning into a data‑driven, adaptive process. |
How It Works
- Real‑time demand monitoring – An AI agent continuously scans traffic across cells.
- Decision making – The agent decides when and how to admit new users to specific cells.
- Adaptive optimization – As conditions stabilize, the agent refines its policies, converting traditional manual configurations into a self‑optimizing, data‑driven loop.
Result: More resilient, efficient mobile networks that can automatically adjust to changing load patterns without human intervention.
Evolving Network Configuration With Multi‑Agent Orchestration
Telcos can now design, observe, and optimize complex agentic workflows across the RAN thanks to a collaboration between NVIDIA and BubbleRAN. Both companies are extending the NVIDIA Blueprint for Telco Network Configuration with:
These complementary frameworks enable multi‑agent orchestration.
Integration with Opti‑Sphere
BubbleRAN is embedding NAT and BAT into its Opti‑Sphere platform. The integration provides:
- Flexible management of network‑monitoring, configuration, and validation agents across containers and workloads.
- Seamless connection to tools that report network metrics and traffic status.
- Continuous proposal and validation of configuration changes.
First Deployment
Telenor Group will be the inaugural telco to adopt this blueprint with BubbleRAN, enhancing its 5G network for Telenor Maritime, the group’s global connectivity provider at sea.
Learn More
- Mobile World Congress – Discover the latest advancements in agentic AI for telecommunications.
- Date: March 2‑5, 2026
- Location: Barcelona
- Notice – See the terms of service for software product information.