Nvidia has partnered with global telecommunications operators to develop telecom AI agents designed to transition networks from task-based automation to autonomous operations.
The collaboration includes SoftBank Corp and NTT Data. These companies plan to demonstrate a stack of data, models, simulation tools, and secure runtimes at the DTW Ignite 2026 conference.
While generative artificial intelligence has delivered returns in network management, its impact remains limited. Currently, humans must still correlate insights and direct next actions. Consequently, Nvidia argues that automation should serve as a launchpad to full autonomy.
Transitioning to Autonomous Networks
Under the new framework, autonomous systems will proactively monitor for problems. Meanwhile, they will coordinate changes across network, IT, and business systems. This approach ensures that human operators still maintain control over high-level policies.
Several partners are already building on this framework. Specifically, AdaptKey is piloting self-healing 5G agents with operators. In addition, ServiceNow is bringing its incident-response tooling to network operations centres, while Tata Consultancy Services is developing a sensor architecture for anomaly detection.
Addressing Data Privacy Challenges
Data availability remains a significant obstacle for the telecommunications sector. Notably, Nvidia cited figures showing that 54% of operators view data-related issues as their biggest obstacle to building industry-specific models. This difficulty arises because valuable network and customer data is too sensitive to use directly for training.
To address this, SoftBank Corp is using Nvidia’s NeMo Safe Synthesizer and NeMo Anonymizer tools. These tools generate privacy-preserving synthetic datasets that mirror real network performance. As a result, the company can fine-tune its Large Telecom Model without exposing sensitive user information.
Deploying Telecom AI Agents in Runtimes
The deployment of telecom AI agents requires secure environments to prevent unauthorized actions. To achieve this, Nvidia designed its NemoClaw blueprints and OpenShell secure runtime. These tools provide policy-based guardrails and sandboxed access to telecom systems.
Meanwhile, NTT Data is applying open Nemotron models alongside NemoClaw. This integration allows their systems to track long-term network performance trends. Furthermore, the system escalates anomalies to specialized research agents for deeper telemetry analysis.
Simulating Future Network Conditions
Simulation serves as the third pillar of this collaborative effort. It allows telecom AI agents to test recommendations in digital environments before acting on live systems. Consequently, operators can avoid service disruptions during updates.
Software company Forsk stated its GPU-accelerated radio propagation model achieves high accuracy up to 200 times faster than CPU-only systems. Similarly, Viavi Solutions reports major gains in simulation throughput using Nvidia RTX Pro 6000 Blackwell GPUs. Finally, KDDI is working with Keysight and Samsung Research America to build a high-fidelity digital twin for the 6G era.