Telecom Operators Rely Increasingly on NVIDIA's Open AI Models for Trustworthy, Customizable Solutions

Image: NVIDIA · Source

Telecom companies are adopting open AI models, like NVIDIA's Nemotron series, to enhance control, customization, and trustworthiness in critical operations such as network management and customer service.

Telecom operators worldwide are increasingly basing their artificial intelligence strategies on open models, moving beyond cost considerations to benefit from greater trust, control, and customization. According to NVIDIA’s latest State of AI in Telecommunications report, 89% of surveyed telecom firms consider open source AI models vital to their AI approach.

Open models provide multiple strategic advantages for telecom providers. They enable access to advanced AI capabilities at reduced costs, allowing the selective use of closed models where utmost value exists. Independent benchmarks confirm that state-of-the-art open models now closely match closed alternatives in complex reasoning, coding, scientific, and agentic tasks.

Additionally, open models allow telecom companies to fine-tune AI systems incorporating their unique network, customer, and industry data. This flexibility supports trustworthy AI by granting operators full visibility and governance over model behavior to ensure regulatory compliance and business policy alignment. Open AI also facilitates secure and scalable deployments across public clouds, private data centers, and edge environments.

NVIDIA has developed the Nemotron family of open AI models, optimized for reasoning and speech tasks relevant to telecom workflows, and available with open weights and training recipes. For example, SoftBank Corp. utilizes these models to build its proprietary SoftBank Large Telecom Model, continuously enhancing capabilities related to network operations, design, and management.

Furthering this ecosystem, NVIDIA recently launched the Nemotron Coalition, a global partnership among leading AI labs and companies to collaboratively develop frontier open foundation models. This coalition includes innovators such as Mistral AI, LangChain, and Black Forest Labs and aims to accelerate open model progress, culminating in the upcoming Nemotron 4 model family.

To enable practical implementation, NVIDIA provides end-to-end software and toolkits like NVIDIA NeMo, which aids telecoms in fine-tuning open models with their data while ensuring privacy and governance. AT&T highlights the importance of matching AI workloads to the appropriate models for optimal performance, cost efficiency, and control, leveraging NVIDIA's platform to operationalize diverse AI solutions securely.

Telecom operators are also tailoring AI services to local languages, regulations, and cultural needs. Indonesian operator Indosat Ooredoo Hutchison, for instance, uses open models via its Sahabat-AI project to create AI systems attuned to Indonesian languages and contexts.

This growing adoption of open AI models in telecom underscores a shift toward more transparent, customizable, and sovereign AI infrastructures, supporting not only operational efficiency but also localized innovation across the industry.