Barcelona – The future of telecommunications networks is shifting from automation to true autonomy, and NVIDIA is laying the groundwork with a suite of new artificial intelligence tools unveiled ahead of Mobile World Congress. These advancements, detailed today, aim to provide networks the ability to not just execute tasks, but to understand intent, reason through complex scenarios, and self-manage – a critical step toward more resilient and efficient 5G and future networks.
The core of this push is a new open-source large telco model (LTM) based on NVIDIA Nemotron, designed to understand the specific language and workflows of the telecom industry. This isn’t simply about automating existing processes; it’s about enabling networks to make informed decisions, adapt to changing conditions, and proactively address issues. According to NVIDIA’s recent State of AI in Telecommunications report, network automation is currently the top investment priority for telcos, but the industry is rapidly recognizing the need to move beyond simple automation toward full autonomy.
The move toward autonomous networks is driven by the increasing complexity of modern telecommunications infrastructure. Operators are facing challenges in managing diverse, multi-vendor environments, optimizing energy consumption, and ensuring consistent quality of service. AI, and specifically agentic AI, offers a path to address these challenges by providing a layer of intelligent control that can adapt and respond in real-time.
Building Networks That Can Consider for Themselves
The NVIDIA Nemotron LTM, developed in collaboration with AdaptKey AI, is a 30-billion-parameter model specifically fine-tuned on telecom datasets. This specialization allows it to understand industry terminology and reason through complex tasks like fault isolation, remediation planning, and change validation. Crucially, the model is open source, giving telcos transparency into its training and data, and enabling secure, on-premises deployment. This addresses concerns about data security and control, allowing operators to build and run AI agents directly within their own networks.
But a powerful model is only one piece of the puzzle. NVIDIA is similarly releasing a guide, co-authored with Tech Mahindra, detailing how to fine-tune these reasoning models and build agents capable of safely executing network operations center (NOC) workflows. The guide emphasizes a framework for teaching AI to “reason like a network engineer,” focusing on high-impact incidents and translating expert resolutions into structured, step-by-step procedures. This approach aims to create AI agents that not only know *what* to do, but *why* a particular course of action is effective and safe.
Energy Efficiency and Network Configuration
Beyond reasoning and problem-solving, NVIDIA is also tackling practical challenges like energy consumption. The company unveiled a new Blueprint for intent-driven RAN energy efficiency, integrating NVIDIA technology with VIAVI’s TeraVM AI RAN Scenario Generator (AI RSG). This blueprint allows operators to simulate energy-saving policies in a closed-loop system, validating their effectiveness without disrupting live networks. VIAVI’s platform generates synthetic network data, allowing the AI to propose and test energy-saving measures before implementation.
NVIDIA’s Blueprints are also being applied to network configuration. Cassava Technologies is leveraging the NVIDIA Blueprint for telco network configuration to build Cassava Autonomous Network, a platform designed to optimize its network across Africa. The platform utilizes three AI agents: one for monitoring and recommending changes, one for applying those changes with built-in governance, and one for assessing impact and rolling back changes if necessary. NTT DATA is also implementing the blueprint in Japan, using AI to intelligently manage traffic surges and improve network resilience.
Multi-Agent Orchestration and Future Expansion
To manage the complexity of these agentic workflows, NVIDIA is collaborating with BubbleRAN to enhance the Blueprint with the NVIDIA NeMo Agent Toolkit (NAT) and BubbleRAN Agentic Toolkit (BAT). These frameworks provide tools for designing, observing, and optimizing multi-agent orchestration across the RAN. BubbleRAN’s Opti-Sphere platform will integrate these tools, allowing for more flexible management of network monitoring, configuration, and validation agents. Telenor Group is set to be the first telco to adopt this enhanced blueprint, focusing on improving its 5G network for Telenor Maritime.
These advancements are being released as part of GSMA’s new Open Telco AI initiative, signaling a broader industry effort to embrace open-source AI solutions for telecommunications. The goal is to provide a common set of building blocks for the mobile ecosystem, fostering innovation and accelerating the adoption of autonomous networks.
The rollout of these tools and blueprints comes at a pivotal moment for the telecommunications industry. As networks become increasingly complex and demand for bandwidth continues to grow, the need for intelligent, self-managing systems is becoming more urgent. The next step will be seeing these technologies deployed at scale and observing their impact on network performance, efficiency, and reliability. Further demonstrations and discussions on these advancements will take place at Mobile World Congress in Barcelona, March 2-5.
Share your thoughts on the future of autonomous networks in the comments below.
