NVIDIA GTC Paris: Building the Future with AI-Native Telecom Networks

In the rapidly evolving world of telecommunications, the integration of artificial intelligence (AI) is no longer a distant vision—it's happening now. At the recent NVIDIA GTC Paris event, I had the privilege of sharing insights on how telecom companies, particularly in Europe, are pioneering the development of AI-native networks that are set to revolutionize the industry. This transformation is not just about faster connectivity but about embedding intelligence directly into the fabric of telecom infrastructure to unlock new business opportunities, enhance customer experiences, and enable autonomous network operations.

In this article, I’ll take you through the journey of building an AI-native telco—from the foundational infrastructure to advanced AI-powered network operations and the exciting future of 6G networks. Along the way, I’ll highlight real-world examples of telecom operators who are already leading the charge and explain why the future of telecom is inseparable from AI.

🌐 The Vision of an AI-Native Telco

What does it mean to be an AI-native telco? Simply put, it’s a telecommunications provider that leverages AI not just as an add-on but as a core part of its operations, customer service, and innovation strategy. AI-native telcos use AI to:

• Optimize and automate network operations
• Enhance customer experience through intelligent agents
• Accelerate innovation and create new revenue streams

For many in the telecom industry, innovation has often been hampered by the need to maintain and operate complex networks simultaneously. AI offers a way to break through these constraints by automating routine tasks, predicting and resolving network issues before they impact users, and enabling new services that were previously unimaginable.

At NVIDIA, we see this era as incredibly exciting. Today, over 150 telecom operators around the globe—and 90% of the top 50 telcos—are collaborating with us to bring AI into their networks. In fact, a recent survey of 450 telecom professionals revealed that 97% are already using or planning to use AI in some capacity. Clearly, the industry recognizes that AI is not optional; it’s the future.

🤖 Why AI is Essential for Future Telecom Networks

To understand the urgency and scale of this transformation, consider the following questions:

• Will we soon be able to talk to AI agents on our phones, at scale, with millions of simultaneous conversations?
• Will our cities become safer through AI-powered video traffic analysis, delivering real-time insights on congestion and incidents?
• Will autonomous vehicles, like robo-taxis, rely on ultra-reliable, real-time network connections to navigate our streets?
• Will public and private spaces—from airports to manufacturing plants—depend on fleets of connected drones and robots that require constant retraining and updates?

If you answered yes to any of these, you understand the massive connectivity and intelligence demands that lie ahead. Millions, if not billions, of real-time or near-real-time connections will be required, each demanding not only bandwidth but also embedded intelligence to make sense of the data and act autonomously.

This is why telecom networks must evolve from being mere “connectivity fabrics” to becoming “intelligence fabrics.” It’s not enough to connect devices; networks need to embed compute power and AI capabilities at every level to deliver the services of tomorrow. This evolution hinges on the development of new “muscles” within telcos, including:

• Sovereignty: Telcos are trusted providers of critical infrastructure in their regions, making data sovereignty a fundamental requirement.
• Determinism: Real-time applications require deterministic compute and network cycles to guarantee performance.
• Ecosystem Development: Building an AI-native telco requires a robust ecosystem of partners and software providers.
• Homogeneous Infrastructure: A flexible, scalable hardware and software foundation that can support diverse workloads and be upgraded seamlessly.

These capabilities form the foundation of what it takes to become an AI-native telco.

🏗️ Building Block 1: Homogeneous, Scalable, Accelerated Infrastructure

At the core of an AI-native telco is a homogeneous infrastructure—a unified hardware and software environment that can be dynamically provisioned to support a wide range of AI and networking workloads. This infrastructure must be scalable, accelerated (leveraging GPUs and AI accelerators), and flexible enough to run everything from radio access network (RAN) functions to AI applications.

While hardware is the base, it’s really about the integration of real estate, power, and compute resources that create a foundation for AI workloads. On top of this lies a disaggregated software layer that allows telcos to run diverse applications with ease. This includes everything from traditional RAN functions to the advanced AI libraries and models that power new services.

Many telcos are already making this transition. For example:

• Swisscom: Launched their Swiss AI platform with GPU-as-a-service in late 2022, evolving it into an AI work hub for developing AI applications.
• Telenor: Partnered with Capgemini to deliver instant translation services in over 100 languages, with a critical use case for the Red Cross.
• Fastweb: Released version two of their Italian language AI model, Mia, accelerating AI adoption in their market.
• Orange: Transitioned their live intelligence platform from internal use to an open market offering, demonstrating learn-by-doing in AI deployment.
• STC Solutions: Joined the NVIDIA Cloud Partner Program to deliver GPU-as-a-service and platform-as-a-service offerings.
• Telefonica: Deploying distributed intelligence infrastructure that supports data sovereignty and digital initiatives in Spain.

These examples illustrate how telcos are not only modernizing their infrastructure but also creating new revenue streams through AI-driven services, moving beyond traditional average revenue per user (ARPU) models to token-based monetization.

⚙️ Building Block 2: Autonomous Operations Powered by AI Agents

Once the infrastructure is in place, the next critical step is enabling autonomous operations. This means using AI to automate network management, customer service, and employee workflows, reducing operational costs and improving service quality.

Large language models (LLMs) and AI reasoning capabilities are at the heart of this transformation. AI agents—autonomous software entities that can reason, plan, and execute tasks—are increasingly being deployed to handle complex network operations.

Our ecosystem partners have developed numerous AI-driven network operation solutions that demonstrate impressive results:

• 2x return on investment (ROI)
• 22% capital savings
• 63% reduction in average call handling time
• Significant decrease in mean time to resolution for network issues

To support telcos in adopting these innovations, NVIDIA recently released its first AI blueprint for network operations. This blueprint provides network engineers with documentation, reference architectures, and ready-to-use, high-performance AI models that enable real-time, dynamic network parameter configuration through agentic flows.

Agentic flows represent a hierarchy of AI agents working collaboratively—super agents overseeing worker agents—enabled by digital twin technology for visualization and simulation. This approach simplifies what was once a complex, manual process, allowing networks to adapt instantly to changing demands.

Some notable AI-native telco solutions from our partners include:

• Accenture: Network operation center AI app that manages ticket deflection and alarm suppression using hierarchical agents.
• TCS: AI-native telco suite addressing network operations, customer experience, and IT functions with advanced agents leveraging large language models.
• Infosys: Smart network assurance agent employing reasoning models to provide robust guardrails and network management.
• NTT DATA: End-to-end autonomous network alarm management solution.
• Prodapt: Agent squad system with multiple personas and agents collaborating to support engineers in the field and network operations center.

These examples show how AI is enabling telcos to transform network operations from reactive to proactive and autonomous, unlocking new efficiencies and service quality improvements.

📡 Building Block 3: AI-Native Radio Access Networks and 6G Evolution

The radio access network (RAN) is the critical last mile of connectivity. Traditionally, RANs have relied on proprietary, single-purpose hardware and static algorithms for signal processing. However, the future demands a radical shift—AI-native RANs that fuse AI with wireless infrastructure to create dynamic, adaptable, and intelligent networks.

We’re already seeing this with AI RAN, where neural networks replace traditional signal processing algorithms to improve spectral efficiency, throughput, and coverage. This evolution is critical as we prepare for 6G, whose standards are currently being defined with AI as a foundational element.

Key innovations in AI-native RAN include:

• Integrated sensing: Using the radio network as a sensor to dynamically adapt to environmental factors such as weather, terrain, and user density.
• Semantic communications: Transmitting only the essential information using AI to infer meaning, dramatically reducing bandwidth requirements.
• Non-terrestrial networks: Integrating satellite and aerial networks as overlays to cellular networks for extended coverage and resilience.

To illustrate this, NVIDIA developed and demonstrated AI aerial technology that trains, simulates, and deploys neural network models to replace traditional 5G channel estimation algorithms. The results are striking:

• Nearly double the cell throughput in simulations
• 100% higher throughput in live 5G network deployments
• Continuous learning and improvement through digital twin feedback loops

Developers can now go from code to a working AI model in a live 5G network in just a few hours using the NVIDIA Shona research kit and aerial radio frameworks. This democratization of AI RAN development is accelerating innovation at an unprecedented pace.

Several telcos and research institutions are embracing AI RAN and 6G development:

• Indonesia’s IOH: Leveraging AI RAN to deliver education and health services across a vast archipelago.
• T-Mobile US: Integrating AI RAN in mobile switching offices to enhance network performance.
• SoftBank Japan: Conducting field trials demonstrating the importance of low-latency AI at the edge.
• Europe’s Research Community: Over 200 institutions across 33 countries using NVIDIA tools to pioneer 6G research.

This wave of innovation is supported by the AI RAN Alliance, which has grown from 10 to 75 member companies in just over a year, illustrating the rapid industry-wide commitment to AI-native wireless networks.

🧠 Building Block 4: Digital Twins for Simulation and Network Visualization

Digital twins are virtual replicas of physical systems that allow for precise simulation and visualization. In telecom, they enable accurate modeling of radio frequency (RF) environments, network operations, and AI agent activity.

Because wireless signals behave according to physical laws—reflecting, refracting, and being absorbed by materials like concrete, glass, and foliage—digital twins must simulate these phenomena with high fidelity. This allows operators to test network configurations, AI models, and operational strategies in a virtual environment before deploying them live.

Digital twins also provide a powerful interface for network operations by visualizing real-time network status and AI agent interactions. This fusion of AI, network data, and visualization tools empowers operators to make informed decisions and respond swiftly to issues.

Together with the other building blocks, digital twins complete the template of an AI-native telco:

• Homogeneous infrastructure for flexible compute and AI workloads
• AI-powered autonomous operations for network management and customer experience
• AI-native RAN and 6G networks for dynamic, scalable connectivity
• Digital twins for simulation, testing, and visualization

This integrated approach is already delivering new efficiencies and capabilities at scale, proving that the AI-native telco is not a distant dream but a present reality.

🚀 The Journey Forward: When to Join the AI-Native Telco Revolution

The transformation into an AI-native telco is accelerating rapidly. Many operators worldwide have embarked on this journey, building the necessary infrastructure, adopting AI-powered network operations, and experimenting with AI-native wireless technologies.

The question is no longer if you should adopt AI in telecom but when and how. Early adopters are already seeing tangible benefits—new revenue streams, improved operational efficiency, enhanced customer satisfaction, and the ability to deliver innovative services.

NVIDIA is proud to be at the forefront of this transformation, providing the hardware, software, AI models, and ecosystem partnerships that empower telcos to succeed. Whether it’s through our Cloud Partner Program, AI blueprints, or developer tools like the Shona research kit, we are committed to democratizing AI capabilities for the telecom industry.

For telecom operators, the time to act is now. The destination of an AI-native telco is within reach, and the path forward is clear. By embracing AI as a native capability, telcos can unlock unprecedented value for consumers, enterprises, and nations alike.

Thank you for joining me on this exciting journey into the future of telecommunications. The next chapter in connectivity is here, and it’s powered by AI.