Telecom Operators Embrace Open AI Models for Customization and Control
Telecom operators are increasingly building their artificial intelligence (AI) strategies on open models to enhance trust, control, and customization across critical workloads, including autonomous networks and customer care. This marks a strategic shift from relying solely on proprietary AI solutions.
Evidence and context
According to NVIDIA’s State of AI in Telecommunications report, released on October 6, 2026, 89% of surveyed telecom companies acknowledged the importance of open-source models and software in their AI strategies. These models allow operators to fine-tune AI for domain-specific tasks using proprietary network and customer data, while also addressing privacy and compliance concerns. The report highlights that open models are not just cost-effective but also offer greater operational flexibility.
Key industry players like SoftBank Corp. and Indosat Ooredoo Hutchison are leveraging open AI frameworks to develop telecom-specific models. SoftBank’s Large Telecom Model, for instance, builds on NVIDIA’s Nemotron framework to advance network operations and management. Similarly, Indosat’s Sahabat-AI family adapts AI to local languages and cultural contexts in Indonesia, demonstrating the role of open models in enabling localized innovation.
In addition to customization, open models provide operators with operational control and governance. For example, AT&T has emphasized the importance of aligning AI workflows with business priorities. Andy Markus, the company’s chief data and AI officer, stated that "open models are essential" for implementing a flexible and governed approach to AI. AT&T’s OTel 2.0 model, built using 400 billion telecom-specific tokens, exemplifies how open frameworks can be tailored for industry-specific tasks.
NVIDIA plays a pivotal role in facilitating this shift through its Nemotron family of open models and supporting tools like the NVIDIA NeMo libraries. These resources enable telecom operators to adapt open models for tasks such as network configuration, customer incident triage, and autonomous operations. Additionally, the recent launch of the Nemotron 3 Large Telco Model, fine-tuned on telecom datasets, promises improved accuracy for industry-specific use cases.
Why it matters
The move toward open AI models reflects telecom operators’ need for solutions that can be customized and governed while reducing dependence on proprietary vendors. This hybrid approach allows operators to balance cost, performance, and control. While proprietary models may still dominate certain real-time applications, open models are increasingly preferred for tasks requiring domain-specific adaptation and compliance with privacy regulations.
As national AI strategies emphasize localized innovation, open models also enable telecom companies to deliver tailored services to diverse markets. For instance, Indosat’s Sahabat-AI illustrates how open AI allows countries like Indonesia to develop AI solutions that align with local languages and cultural requirements.
The adoption of open AI models underscores a broader trend in the telecommunications industry: leveraging flexible and transparent technologies to meet evolving operational and regulatory needs. With continued advancements in open AI frameworks, telecom operators are well-positioned to scale these solutions into production workflows, driving efficiency and innovation across the sector.