NVIDIA has announced at GTC that its partners are developing large telco models (LTMs) and AI agents designed for the telecoms industry.

Global telecoms networks handle an immense volume of data, supporting millions of user connections daily and generating an average of over 3,800 terabytes of data every single minute. This massive data flow – encompassing network traffic information, performance metrics, configuration details, and topology – is inherently unstructured and complex.

Traditional automation tools have often struggled to manage such substantial, real-time workloads. New solutions are being developed to tackle this historic issue using NIM and NeMo microservices within the NVIDIA AI Enterprise software platform. These LTMs and AI agents aim to usher in the next era of AI in network operations.

LTMs – which are customised, multimodal large language models (LLMs) trained on specific telecoms network data – form a crucial foundation for the development of network AI agents. These agents will automate intricate decision-making processes, enhance operational efficiency, boost employee productivity, and ultimately improve overall network performance.

Companies like SoftBank and Tech Mahindra have already developed new LTMs and AI agents. Meanwhile, Amdocs, BubbleRAN, and ServiceNow are also enhancing their network operations and optimisation capabilities with the introduction of new AI agents, all leveraging NVIDIA AI Enterprise.

According to a recent NVIDIA-conducted telecom survey, 40% of respondents are in the process of deploying AI into their network planning and operational frameworks.

LTMs: Understanding the language of networks

Just as large language models (LLMs) excel at understanding and generating human language, LTMs now enable AI agents to effectively master the “language” of telecom networks.

The newly-developed LTMs offer several key advantages:

• Specialised network intelligence: These LTMs possess the ability to comprehend real-time network events, proactively predict potential failures, and automatically implement resolutions.
• Optimised for telecom workloads: By leveraging NVIDIA NIM microservices, the LTMs are optimised for efficiency, accuracy, and low latency, crucial for demanding telecoms applications.
• Continuous learning and adaptation: With post-training scalability facilitated by NVIDIA NeMo, the LTMs can learn from new events, alerts, and anomalies to continuously enhance their performance over time.

The NVIDIA AI Enterprise platform provides a suite of additional tools and blueprints to facilitate the creation of AI agents. These agents are designed to simplify network operations, deliver significant cost savings and improved operational efficiency, all while enhancing critical network key performance indicators (KPIs) such as:

• Reduced downtime: AI agents can predict failures before they occur, leading to greater network resilience and minimising disruptions.
• Improved customer experiences: AI-driven optimisations result in faster network speeds, fewer outages, and a more seamless connectivity experience for users.
• Enhanced security: By continuously scanning for potential threats, AI can play a vital role in mitigating cyber risks in real-time to bolster network security.

Telecoms giants pioneer LTMs and AI agents with NVIDIA

Across the telecoms industry, leading companies are embracing NVIDIA AI Enterprise to drive advancements in their latest technologies.

SoftBank, for example, has engineered a novel LTM based on a large-scale LLM base model that has been trained using its proprietary network data. Initially focused on network configuration, this model – now available as an NVIDIA NIM microservice – can automatically reconfigure the network to adapt to fluctuations in network traffic, particularly during high-demand events at stadiums and other large venues.

Tech Mahindra has developed an LTM utilising NVIDIA’s agentic AI tools to address critical aspects of network operations. Leveraging this LTM, the company’s Adaptive Network Insights Studio provides a 360-degree view of network issues. It generates automated reports with varying levels of detail, providing insights and assistance to IT teams, network engineers, and company executives.

Furthermore, Tech Mahindra’s Proactive Network Anomaly Resolution Hub is powered by the LTM to automatically resolve a significant proportion of network events. This automation reduces the workload on engineers and enhances their overall productivity.

Amdocs’ Network Assurance Agent, powered by their amAIz Agents framework, automates repetitive yet crucial tasks such as fault prediction. It also conducts in-depth impact analysis and suggests preventative measures for potential network issues, providing step-by-step guidance on resolving any problems that do arise. Additionally, their Network Deployment Agent simplifies the adoption of open radio access network (RAN) technology by automating integration, deployment tasks, and interoperability testing, while also providing valuable insights to network engineers.

BubbleRAN is in the process of developing an autonomous multi-agent RAN intelligence platform built on a cloud-native infrastructure. This platform will utilise LTMs to observe the network’s state, configuration, availability, and key performance indicators (KPIs) to streamline monitoring and troubleshooting processes.

ServiceNow’s AI agents in the telecoms sector, built with NVIDIA AI Enterprise on NVIDIA DGX Cloud, are designed to boost productivity by automatically generating resolution playbooks and predicting potential network disruptions before they occur. This proactive approach helps communications service providers significantly reduce resolution times and improve overall customer satisfaction. These new, ready-to-use AI agents also analyse network incidents to identify the underlying root causes of disruptions, enabling faster resolution and preventing future occurrences.

NVIDIA’s collaborative efforts with its partners are ushering in a new era of intelligent telecoms networks. As the demand for data continues to surge, these AI-driven innovations will be crucial in enabling telcos to manage increasingly complex networks.

(Image credit: NVIDIA)

See also: Alex Leadbeater, GSMA: Security collaboration vital as attack surface grows

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.