Nvidia released an open source Large Telco Model aimed at enterprises in the telecommunications industry looking for AI models that are trained specifically on their data and knowledgeable about their processes.
The AI hardware-software giant said it built the new model on its Nemotron 3 family of foundation models, which it released in December.
The Large Telco Model (LTM) is trained to understand the telecommunications industry’s language and processes through workflows such as fault isolation, the process of locating the part of a system that is responsible for failure; remediation planning, or planning ways to correct a deficiency or failure; and change validation, the process of approving changes to a system, Nvidia said. Aside from the LTM, Nvidia unveiled its Intent-Driven RAN Energy Efficiency Blueprint, a closed-loop agentic workflow for energy optimization.
The LTM, introduced on Feb. 28 at the Mobile World Congress Barcelona, reflects a trend in which telecommunications companies increasingly require domain-specific models to drive more autonomous networks. Nvidia is not the only model provider working toward providing these domain-specific models. For example, in partnership with Vodafone, Microsoft is deploying Azure-powered agents that can be used for network operations. AMD is also participating in the “Open Telco AI” initiative and providing hardware and compute power to run telco-specific models.
Nvidia’s LTM for Telcos
For its part, with the open source LTM, Nvidia is enabling telcos to create their own datasets and integrate them more deeply into their systems and networks.
“[It’s] a model that’s been trained and developed using all the industry standards and information and data sets,” said Susan Welsh de Grimaldo, an analyst at Gartner.
Beyond providing a domain-specific model, Nvidia is also offering a way to address the challenges telco teams face when automating their processes, said Nick Patience, an analyst at Futurum Group.
“Current automation is rules-based,” Patience said. “It breaks down the moment something falls outside the script.”
The LTM focuses on the layer above the system that can interpret the operator’s intent and make decisions that weren’t explicitly stated, using its reasoning mode, he said.
“Network operations have real consequences when they fail, and reasoning models work through multi-step problems rather than pattern-matching to a likely answer,” Patience said.
Nvidia is also focusing on transparency, security, guidelines, and governance for its AI models, recognizing that these are important to enterprises in the telco market, he said.
“When you think about autonomous networks and building AI agents for running networks, we’re going to see both human and AI involvement,” Welsh de Grimaldo said. She added that Nvidia, understanding this, realized that the models need to be trained to understand skills building from a network operations engineer’s perspective, which can “drive better improvements in how it’s used in running different functions inside the network.”
Some Obstacles
However, a key challenge Nvidia faces is that it competes with traditional network vendors with well-established customer bases, such as Ericsson and Nokia.
“Nvidia’s new to the domain,” said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget. He added that while the AI hardware provider has been working with telco vendors for years, this is the first time the vendor is offering its product as a competing alternative. Moreover, AI is not the only answer to some of the challenges in the telco arena.
“Bringing AI to the table does help reduce a lot of the complexity, but fundamentally, there are a lot of more nuanced challenges that Nvidia alone won’t be able to solve,” Su said.
Another challenge is that while the model is open source, it remains to be seen “whether telco IT organizations can absorb it fast enough to matter,” Patience said.
