Global Release — In a profound strategic pivot, the world’s leading telecommunications operators are increasingly bypassing closed, proprietary systems to anchor their artificial intelligence futures in open models. This shift extends far beyond immediate cost efficiencies. By embracing open-source foundations, architectures, and software, major global operators are gaining unprecedented levels of trust, data control, and deep operational customization. These open models are rapidly becoming the bedrock for the industry’s most critical, mission-critical workloads, ranging from self-optimizing autonomous networks to hyper-responsive customer care ecosystems.

NVIDIA’s landmark State of AI in Telecommunications report captures the sheer scale of this industry-wide transition. According to the data, an overwhelming 89% of telecommunications respondents consider open-source models and software to be critically important to their overarching corporate AI strategy. As networks grow increasingly complex with the rollout of 5G Advanced and the horizon of 6G, operators are finding that rigid, black-box AI tools cannot keep pace with the nuanced, high-stakes demands of modern telecommunications infrastructure.

Main Facts: The Strategic Pivot to Open AI in Telecom

The adoption of open models across the telecom sector is driven by a fundamental need for architectural flexibility and sovereignty. Unlike proprietary software ecosystems that lock enterprises into single-vendor trajectories, open models provide carriers with the keys to their own technological destiny.

At the center of this movement is the NVIDIA Nemotron family of open models, which delivers frontier-level reasoning performance specifically optimized for agentic workflows, complex voice applications, and edge processing. Crucially, these models provide open weights, transparent training data, and explicit execution recipes.

To bridge the gap between general-purpose AI and the highly specialized lexicon of telecommunications, industry collaboration has yielded domain-specific breakthroughs. For instance, NVIDIA announced the 30-billion-parameter Nemotron 3 Large Telco Model (LTM). Fine-tuned by AdaptKey using extensive open-source telecom datasets, the LTM delivers unprecedented accuracy gains for specialized tasks. It equips operators with an open baseline that inherently understands specialized industry terminology, allowing it to reason effectively through intricate telecom workflows such as dynamic network configuration, automated incident triage, and predictive maintenance.

Furthermore, to ensure operators are not left guessing how to adapt these architectures, NVIDIA released a complete end-to-end fine-tuning pipeline. Leveraging NVIDIA NeMo open libraries, this framework guides engineers step-by-step through the process of tailoring open models to proprietary operator networks, customer bases, and internal operational procedures.

Chronology of Evolution: From Closed Experiments to Open Production

The journey of artificial intelligence within the telecommunications sector has undergone a rapid, multi-stage evolution over the past half-decade.

• Phase 1: The Era of Siloed PoCs (2018–2021): Early AI adoption in telecom was largely characterized by isolated proof-of-concepts (PoCs). Operators experimented with predictive analytics and basic customer service chatbots, typically relying on closed, vendor-managed cloud APIs. These early implementations were hindered by data privacy concerns, latency issues, and an inability to customize models for deep network operations.
• Phase 2: The Rise of Foundation Models and Cost Realities (2022–2023): The explosion of generative AI introduced powerful frontier models. However, telcos quickly realized that sending sensitive subscriber data and proprietary network telemetry to external, closed-model providers introduced unacceptable security and compliance risks. The financial overhead of scaling proprietary APIs across millions of daily transactions also proved unsustainable.
• Phase 3: The Open Model Awakening and Domain Specialization (2024–2025): Recognizing the need for local control, operators began championing open-source architectures. Tech giants and specialized startups alike began releasing open weights, enabling carriers to safely ingest internal network datasets. Specialized fine-tuning emerged, turning general models into telecom-fluent engines capable of understanding cellular protocols and switching logic.
• Phase 4: Production-Scale Agentic Workflows and Sovereign AI (2026 and Beyond): Today, the industry has crossed into the era of scaled production. Open models are no longer just experimental sandboxes; they are the core engines powering autonomous networks, national sovereign AI initiatives, and multi-layered agentic ecosystems orchestrated by enterprise-grade software stacks.

Supporting Data and Industry Metrics

The qualitative shift toward open models is powerfully reinforced by hard metrics across the telecommunications landscape.

• 89% Adoption Sentiment: As highlighted in NVIDIA’s latest industry report, nearly 9 out of 10 telecom leaders view open-source software and models as vital components of their institutional AI roadmaps.
• 30 Billion Parameters: The scale of domain-optimized intelligence is exemplified by models like the Nemotron 3 Large Telco Model, which packs 30 billion parameters fine-tuned explicitly for telecom reasoning, striking an optimal balance between computational efficiency and high-level cognitive capability.
• Multilayered Ecosystem Integration: Modern deployments rely on unified platforms such as NVIDIA AI Enterprise and the NVIDIA Agent Toolkit, which integrate data anonymization pipelines, synthetic data generation, secure agent runtimes, and real-time simulation environments.

These metrics demonstrate that telecom AI has matured past simple proof-of-concept deployments into high-throughput, security-first production environments where reliability is non-negotiable.

Official Responses and Executive Insights

Industry leaders are speaking out about why open models have become indispensable to their long-term infrastructure strategies.

SoftBank Corp. and the Pursuit of Telco-Specific Intelligence:
SoftBank Corp. serves as a premier example of an operator leveraging open models as a springboard for proprietary innovation.

"Open models allow SoftBank Corp. to build on the rapid progress of global foundation models while applying the network knowledge and operational expertise we have accumulated over many years," stated Rajeev Koodli, principal fellow at SoftBank Corp. and senior vice president of SB Telecom America.

He added: "আমরা are using open foundations extensively, including NVIDIA Nemotron models and others, in developing our SoftBank Large Telecom Model, continuously advancing it for telecom-specific use cases such as network operations, design and overall management."

AT&T and the Multi-Model Production Strategy:
For massive tier-1 carriers like AT&T, the core advantage lies in strategic flexibility and workload optimization.

"At AT&T, we believe the future of AI is not about choosing a single model; it’s about intelligently matching every workload to the right combination of performance, cost and control," noted Andy Markus, chief data and AI officer of AT&T.

Markus emphasized the operational reality of these choices: "Open models are essential to that approach, and with NVIDIA, we can bring that model-choice strategy into production with the scale, reliability and governance our business requires."

Indosat Ooredoo Hutchison and Local Sovereign AI:
Beyond network operations, open models are unlocking localized cultural and linguistic representation. Chirag Sukhadia, chief data and AI officer of Indosat Ooredoo Hutchison, one of Indonesia’s largest operators, detailed the societal and operational impact of localized open initiatives like the Sahabat-AI family of models:

"For countries like Indonesia, the value of open models goes beyond access to powerful AI. It is about adapting that intelligence to our own language, culture, data and real-world needs," Sukhadia explained.

He concluded: "Sahabat-AI puts this into practice, using open models as a foundation to build AI that understands Indonesia and can be developed for local applications. This allows us not only to adopt AI, but to build capabilities around it and enable more Indonesians to create with AI on their own terms."

Implications for the Future of Telecommunications

The widespread migration toward open models carries profound implications for the global telecommunications sector, touching upon security, operational autonomy, and national sovereignty.

1. The Realization of Truly Autonomous Networks

Autonomous networks—systems capable of self-configuration, self-healing, and self-optimization without human intervention—require real-time reasoning engines that understand the physics of signal propagation and core routing logic. By utilizing fine-tuned open models embedded within agentic workflows, operators can automate incident resolution at a speed and scale that legacy rule-based software could never achieve. Network outages can be diagnosed and mitigated in fractions of a second, drastically improving carrier uptime and Quality of Service (QoS).

2. Enhanced Data Governance and Privacy Compliance

Telecom operators handle petabytes of sensitive subscriber metadata, call detail records (CDRs), and location telemetry. Relying on external, black-box cloud models introduces severe compliance liabilities under strict global data protection frameworks like GDPR. Open models allow operators to run inference entirely on-premises or within trusted sovereign cloud environments. Using advanced data pipelines to anonymize sensitive records and generate privacy-preserving synthetic datasets, carriers can train and fine-tune models safely without exposing user data to third parties.

3. Alignment with National Sovereign AI Strategies

As governments worldwide mandate localized control over critical infrastructure, telecommunications providers find themselves at the intersection of national security and digital commerce. Open models provide the ideal vehicle for national operators to build localized AI infrastructure. By adapting models to regional languages, cultural nuances, and strict local regulatory frameworks—as demonstrated by Indosat Ooredoo Hutchison in Indonesia—telcos are positioning themselves as vital pillars of national digital sovereignty.

4. Democratization and Collaborative Innovation

Finally, the embrace of open-source frameworks fosters a vibrant, collaborative ecosystem. Rather than being trapped inside proprietary vendor silos, operators can share methodologies, fine-tuning recipes, and domain-specific datasets through platforms like Hugging Face and GitHub. This collaborative spirit accelerates the pace of innovation, ensuring that even mid-sized regional carriers can harness world-class artificial intelligence tailored precisely to their operational needs.

Ultimately, the transition from closed experimentation to open production marks the coming of age of telecommunications AI. By taking control of their underlying models, operators are safeguarding their networks, empowering their workforces, and laying a resilient, intelligent foundation for the digital economies of tomorrow.