NVIDIA Sets Telecom-Focused AI Push With New Models, Agents

Artificial intelligence chips, hardware, and software maker NVIDIA is pursuing the development of new large telco models (LTMs) and AI agents tailored to the telecommunications industry through an initiative announced by the company at the GTC global AI conference.

The new LTMs and AI agents being custom-built for the telco industry are using NVIDIA NIM and NeMo microservices within the NVIDIA AI Enterprise software platform, with the goal of enabling the next generation of AI in telco network operations.

The company explained that the massive flow of data generated by telecom service providers through their base stations, routers, switches, and data center is now mostly unstructured and complex.

“Not surprisingly, traditional automation tools have often fallen short on handling massive, real-time workloads involving such data,” NVIDIA said.

NVIDIA’s partners in the effort – including SoftBank, Tech Mahindra, Amdocs, BubbleRAN, and ServiceNow – are among the early adopters developing and deploying LTMs and agents with NVIDIA’s enterprise AI tools.

NVIDIA’s new LTMs are custom multimodal large language models trained specifically on carriers’ telco data, enabling a new generation of network AI agents.

These AI agents automate complex decision-making, boost operational efficiency, enhance network performance, and reduce downtime.

LTMs, which are akin to large language models but focused on network intelligence, allow AI agents to interpret real-time network events, predict failures, and trigger automated resolutions.

They leverage NVIDIA NIM microservices for optimized efficiency and latency, while also integrating NeMo to support ongoing learning from fresh alerts and anomalies, NVIDIA said.

NVIDIA’s partners are putting these new capabilities to work:

• SoftBank’s LTM is built on a large-scale LLM and trained on the company’s network data, and automatically reconfigures networks in response to traffic changes – such as spikes during major events. The company is also rolling out agent blueprints to speed up broader telco AI adoption.
• Tech Mahindra’s AI-powered Adaptive Network Insights Studio uses an LTM to offer a 360-degree view of network operations and generate automated reports for engineers and executives. And its Proactive Network Anomaly Resolution Hub employs AI to resolve many network events without human intervention.
• Amdocs has introduced the Network Assurance Agent, part of its amAIz Agents suite, to handle fault prediction, impact analysis, and guided remediation. A separate Deployment Agent supports open RAN by automating integration and testing while supplying real-time insights to engineers.
• BubbleRAN is building an autonomous, cloud-native RAN intelligence platform using LTMs to monitor and troubleshoot network performance. Its AI agents use retrieval-augmented generation pipelines and telco-specific APIs to dynamically answer 5G deployment questions and enforce network policies.
• ServiceNow’s AI agents, running on NVIDIA DGX Cloud, generate playbooks, anticipate disruptions, and trace the root causes of network failures. These tools are designed to cut incident resolution times and improve the end-user experience for telecom customers.