Telecom Operators Have Adopted Open AI Models
Global carriers are shifting toward open-source foundation models to maintain control over network data.
Updated on Oct. 7, 2026 in Artificial Intelligence
Live Poll
Should companies prioritize using open-source software to maintain control over their internal data?
Major telecommunications operators including SoftBank, AT&T, and Indosat Ooredoo Hutchison have begun deploying open-source AI models alongside proprietary systems. This strategy allows carriers to retain control over internal network data while maintaining flexibility in model fine-tuning.
Why it matters
Operators are adopting open-source software to address specific operational requirements that proprietary black-box systems often cannot accommodate. This shift marks a strategic pivot toward custom-tuned models capable of handling localized data and specialized industry tasks.
A 2026 survey indicates that 89% of operators now consider open-source software critical to their AI strategy. These models are increasingly utilized for daily operations, with independent benchmarks suggesting that open foundation models are closing performance gaps in reasoning and coding tasks.
The players
Nvidia
A designer of graphics processing units and AI computing platforms that provides the underlying infrastructure for telecommunications model training.
SoftBank Corp
A major Japanese telecommunications and technology holding company currently integrating internal operational data into its custom Large Telecom Model.
AT&T
A U.S.-based global telecommunications operator incorporating open-source foundation models into its network management strategy.
Indosat Ooredoo Hutchison
An Indonesian telecommunications provider that specializes in adapting open-source AI models to local linguistic and cultural datasets.
The details
Operators are managing hybrid architectures by dividing workloads between proprietary systems for complex, high-stakes tasks and open-source foundation models for daily operations. For example, SoftBank trained its Large Telecom Model using internal operational data on top of the Nvidia Nemotron model family. Indosat Ooredoo Hutchison similarly tailored open models to align with local language and cultural nuances, demonstrating the utility of model fine-tuning for regional relevance.
Timeline
2026: The State of AI in Telecommunications 2026 survey was conducted.
The Tech Race
This move reflects a broader industry competition where telecommunications carriers are attempting to reclaim autonomy from large-scale AI service providers. The shift aligns with research indicating that open foundation models are increasingly competitive against closed-source alternatives in specialized reasoning tasks.
The transition to open models allows operators to process network data with greater speed and localized context. Users may experience improved network management and more responsive customer-facing services as carriers optimize these models for specific regional and technical demands.
The takeaway
Operators are finding that open-source flexibility outweighs the simplicity of proprietary black-box AI systems. Watch for future benchmark reports that compare the efficacy of custom-tuned telecom models against off-the-shelf proprietary solutions.
Further reading
For more on the development of open-source models, see Artificial Intelligence.
Live Poll
Should companies prioritize using open-source software to maintain control over their internal data?
