Why Telecom Operators Are Building Their AI Strategy on Open Models
NVIDIA Blog Kanika Atri
Telecoms are betting on open AI models instead of only closed ones. The draw is control: cheaper, easier to tune, and better for network work.
Based on reporting by NVIDIA Blog, Kanika Atri — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Telecom operators are no longer treating open AI models as a side project. They’re moving them into the center of their strategy, and NVIDIA’s latest State of AI in Telecommunications report says 89% of respondents see open source models and software as important to that plan.
The appeal is broader than price. Open models let operators choose where to use closed systems and where not to, while keeping tighter control over the workloads that matter most: autonomous networks, customer care, and the messy business of telecom operations. They also make customization easier, because operators can fine-tune open weights and training recipes with their own network, customer, and industry data.
Benchmarks are helping the pitch. NVIDIA points to the Artificial Analysis Intelligence Index v4.3.2 as evidence that leading open models are becoming more competitive on reasoning, coding, scientific, and agentic tasks. That matters because telecoms don’t just need chatty demos. They need systems that can understand network terminology, handle customer incident triage, and fit into real operational workflows.
NVIDIA is trying to make that practical with its own open model stack. Its Nemotron family includes open weights, training data, and recipes, and the company also announced the 30-billion-parameter Nemotron 3 Large Telco Model, fine-tuned by AdaptKey on open source telecom datasets. NVIDIA says it released the full recipe for adapting open models through NeMo open libraries, so operators can tune them for their own networks, customers, and procedures.
The last mile is deployment, and that is where telecom gets picky. AT&T says the point is not picking one model but matching each workload to the right mix of performance, cost, and control. NVIDIA says its platform covers data pipelines, model orchestration, secure runtimes, and simulation, while partners help move the whole thing toward production. And for operators like Indosat Ooredoo Hutchison, the promise is local AI in local languages, built around local rules instead of imported assumptions.
My take — AI-written commentary, not fact-checked reporting
Open models are winning in telecom because closed models are a bad fit for a business built on control, regulation, and awkward legacy systems. The funny part is that “open” is now the more practical enterprise choice, while the shiny locked box starts looking like a liability. That pattern is spreading wherever AI has to do real work instead of just impressing a demo room.
Read more about this at: NVIDIA Blog