Details
- NVIDIA says AT&T, SoftBank and Indosat are adopting open models for autonomous network operations and other telecom use cases.
- The company is providing a full software stack to help operators turn open models into specialized AI agents while retaining operational control.
- NVIDIA Nemotron models are designed to reason through network problems, including cases without established runbook procedures.
- NVIDIA OpenShell provides policy-based controls over what agents can access and do, helping operators test and govern automated actions.
- Agents can investigate incidents, compare potential fixes and test remediation plans in network simulations before deployment.
- NVIDIA’s 2026 telecom AI survey found that 89% of respondents consider open-source models and software important to their AI strategy.
- The survey also reported that 66% of respondents are actively using AI, up from 49% in the previous year.
- The announcement positions open, controllable models as a way for telcos to reduce vendor lock-in while developing domain-specific network automation.
Impact
The announcement reflects a broader telecom shift from experimental AI toward controlled automation of network operations. Open models can help operators customize systems for local data, languages and infrastructure, while OpenShell addresses the governance risks of allowing agents to act on production networks. With major operators including AT&T, SoftBank and Indosat involved, NVIDIA is using existing telecom relationships to extend its software and accelerated-computing stack into operational workflows. The approach may pressure competing infrastructure and model vendors to offer similarly specialized, auditable agent tooling rather than generic assistants.
