NVIDIA is taking decisive strides to enhance telecoms network intelligence. At the recent GTC event, the company announced its collaboration with partners on large telco models (LTMs) and AI agents tailored for the telecom industry. These innovations mark a shift towards smarter network management solutions.
Telecom networks handle colossal amounts of data. Every minute, they manage over 3,800 terabytes, highlighting a need for efficient data handling strategies. This volume includes network traffic, configurations, and performance metrics, often in unstructured formats. Traditional tools struggle with this rapid, large-scale data processing. However, NVIDIA aims to change this dynamic using NIM and NeMo microservices within its AI Enterprise software platform. These LTMs, customized for telecoms, form the backbone for AI agents. Such agents could automate complex decision-making, heighten operational efficiency, and upgrade network performance.
Several telecom giants like SoftBank and Tech Mahindra have already adopted new LTMs. Meanwhile, others like Amdocs, BubbleRAN, and ServiceNow are enhancing their operations through AI agents with NVIDIA AI Enterprise.
A survey conducted by NVIDIA shows that 40% of respondents are integrating AI into their networks. LTMs, much like large language models (LLMs), are now decoding the “language” of telecom systems. This offers specialized network intelligence and predicts and resolves failures.
Furthermore, LTMs are optimized for telecom workloads, ensuring efficiency and low latency through NVIDIA NIM microservices. They continuously learn and adapt, improving over time, thanks to NVIDIA NeMo’s post-training scalability. The NVIDIA AI Enterprise platform provides additional resources to create AI agents which streamline operations, reduce downtime, elevate customer experience, and bolster network security.
Notably, SoftBank developed an LTM that autonomously reconfigures networks during high traffic, optimizing performance. Conversely, Tech Mahindra’s Adaptive Network Insights Studio offers detailed network insights, lessening engineer workload and raising productivity. Amdocs and BubbleRAN are also employing AI to automate fault predictions and streamline open RAN deployments. ServiceNow’s AI agents utilize predictive measures to preemptively resolve network disruptions, substantially enhancing customer satisfaction.
