Nvidia Highlights Telecom Operators' Shift to Open-Source AI Models

Image: NVIDIA / official article illustration · Source

Nvidia reveals telecom operators like SoftBank, AT&T, and Indosat Ooredoo Hutchison are adopting open-source AI models to enhance control over network automation, balancing cost, performance, and operational customization.

Nvidia has spotlighted a significant trend among telecommunications operators moving away from exclusive reliance on closed AI systems toward embracing open-source models. This shift aims to grant carriers greater visibility and direct control over network automation and customer workflow management. In a recent blog post, Nvidia outlined how major operators, including SoftBank Corp, AT&T, and Indosat Ooredoo Hutchison, are deploying open models alongside proprietary platforms. These operators benefit from maintaining full oversight of sensitive network data and fine-tuning AI models to meet their unique operational requirements.

This change mirrors a broader transformation across the telecom sector. Nvidia's State of AI in Telecommunications 2026 survey found that 89% of operators view open-source software as critical to their AI strategies. The move is fueled partly by independent benchmarks indicating that open-source models are closing the performance gap with closed systems in areas such as reasoning, coding, and autonomous network issue resolution.

Rather than relying exclusively on a single AI system, operators are partitioning workloads, using costly proprietary models for complex tasks where closed platforms excel, while deploying more cost-effective open foundation models for routine operations. SoftBank, for example, has built its Large Telecom Model on Nvidia’s Nemotron family, leveraging years of its internal operational data. Rajeev Koodli, a principal fellow at SoftBank, emphasized that open models enable the company to harness rapid global AI advancements alongside its accumulated network expertise.

AT&T has adopted a similar multi-model strategy. Andy Markus, its chief data and AI officer, stated that the goal is to strategically match each workload with the optimal balance of performance, cost, and control. Meanwhile, Indosat Ooredoo Hutchison has customized its open-source models to reflect Indonesian language and cultural nuances, illustrating the adaptability of open models to local contexts.

These developments signal a growing preference among telecommunications carriers for AI frameworks that offer transparency, adaptability, and efficiency, balancing innovative technology adoption with operational governance and cost management. Nvidia’s insights provide a clear snapshot of this evolving landscape where open-source AI increasingly underpins network intelligence and automation workflows.