Artificial Intelligence in Telecommunication Networks

🚀 Reflections from #GTC26: The New Era of AI-Native Telecommunications Just wrapped up an incredible week at NVIDIA GTC 2026, and the takeaway is clear: The telecommunications industry is no longer just the "pipe"—it is becoming the backbone of global AI infrastructure. Here are the 3 shifts that redefined the landscape for me: 1️⃣ The "AI Grid" is the New Revenue Engine The NVIDIA AI Grid Reference Design is a game-changer for operators. By transforming existing physical footprints—cell sites, switching offices, and regional hubs—into distributed AI infrastructure, telcos can monetize their edge like never before. It’s no longer just about coverage; it’s about providing high-performance compute exactly where the data is born. 2️⃣ AI-RAN: The Connective Tissue for Physical AI We’ve moved beyond chatbots. Physical AI—systems that perceive, reason, and interact with the physical world—requires near-zero latency to function at scale. AI-RAN is the critical enabler here, providing the low-latency connectivity needed for these autonomous systems to "breathe." In short: If Physical AI is the muscle, AI-RAN is the nervous system. 3️⃣ 6G: Born in the Simulation The road to 6G is being paved in simulation. With the NVIDIA Aerial Omniverse Digital Twin (AODT), we are witnessing the first wireless generation born in a digital environment. Telcos can now architect and validate complex AI-native networks in a risk-free virtual world before a single radio wave hits the live spectrum. #NVIDIAGTC #6G #AIGRID #AIRAN #DigitalTwin #PhysicalAI

Bringing AI Agents to the Heart of 6G Networks! Our latest work: MX-AI – the first end-to-end, multi-agent system that can observe and control a live 5G network through natural language. Future 6G networks will connect billions of devices, demand ultra-low latency, and run on tight energy budgets. Managing this complexity manually is simply impossible. 🤖 MX-AI introduces Large Language Model (LLM)-powered agents that understand operator requests in plain English, monitor the network in real time, and take actions automatically — from diagnosing problems to reconfiguring network slices. What’s unique: ✅ Works on a real 5G Open RAN testbed (not just simulations) ✅ Uses a team of cooperating AI agents (planner, monitor, validator, executor) to turn high-level intentions into precise network commands ✅ Achieves human-expert-level decision accuracy with just seconds of latency ✅ Fully open-sourced to accelerate research on AI-native RANs The result? Operators can ask: “Guarantee 10 Mbps for the emergency services slice from 6–7 pm” …and MX-AI will figure out how to make it happen, implement the change, and confirm it worked. This is a step toward truly intent-driven, AI-native 6G networks — where humans set the goals, and intelligent agents handle the complexity. 📄 Read the paper: https://lnkd.in/dz3muvcN A huge thank you to my amazing co-authors and collaborators: Ilias Chatzistefanidis, Andrea Leone, Ali Yaghoubian, Mikel Irazabal, Sehad Nassim, Lina Bariah, Navid Nikaein. And to our institutions for their support: @EURECOM, @BubbleRAN, @AaltoUniversity, @KhalifaUniversity. #6G #OpenRAN #AI #TelecomInnovation #LLM #MXAI #AgenticAI

Every industry is being reshaped by #AI. But #telecoms? Many still write it off as too slow, too legacy, too regulated. That view is increasingly outdated. Following #MWC2026, I mapped Ericsson's AI deployments across the standard AI stack: Infrastructure, Model, Platform and Application. The picture is more complete than most people realize. This is not a roadmap. Most of it is live today. Here is what stands out: 📡 Infrastructure (Device & RAN) AI embedded at the physical layer. On-device inference, Vehicle-to-Everything (V2X) in connected cars, analytics inside IoT sensors. In the RAN, spectrum optimized continuously, energy cut dynamically, beamforming improved in real time. The network hardware is becoming intelligent. 🧠 Model (Multi-access Edge Computing (MEC)) Where AI models actually run, close to the source with single-digit millisecond latency. Autonomous fault detection, real-time inference, industrial automation, live network simulation. From reactive operations to self-healing behavior. 🏛 Platform / Tooling (5G Core) Orchestration, slicing, policy and APIs all AI-driven. Operators declare intent. AI configures the rest. The role of the network engineer shifts from manual configuration to oversight. ☁️ Application (Cloud & Operations Support Systems (OSS)) AI running operations end-to-end. Predicting failures, automating planning, moving humans to oversight. Federated learning and an AI model marketplace are next. Ericsson is not adding AI to the network. They are rebuilding the network around it. 🔭 Looking ahead Most AI transformations sit on top of infrastructure. In telecoms, it is happening inside it. Near term, cross-layer AI and open Network APIs turn the telecoms stack into a platform others build on. By 2030, 6G makes AI-nativeness a design requirement, not a retrofit. The network stops carrying intelligence and becomes the intelligence layer itself. Telecoms is not catching up to the AI wave. It is becoming the infrastructure the AI wave runs on. Proud to be part of building exactly that at Ericsson. 💡 Which layer of the stack surprises you most? #Telecoms #AI #5G #MWC #Ericsson #NetworkIntelligence #AIStack #6G

Telcos, Stop Renting Intelligence. Start Building Brains. 🧠 As of 2025, fewer than 1% of telecom operators own a foundation AI model. SoftBank, SK Telecom, and China Mobile are examples of Telcos that have trained their own models on their network data. The rest rely on third-party APIs from OpenAI, Google, or Anthropic for analytics, operations, and automation. This dependency will have economic consequences. The problem for Telcos is the cost: A telco-grade foundation model costs about $60–100 million to train once. Renting equivalent intelligence through APIs costs $5–10 million per year, but without data control, model access, or customisation. Over five years, the cost is similar, except that your intelligence is owned by someone else. The alternative is co-training: Several operators share compute and data under an encrypted federation. Each retains sovereignty, yet all benefit from collective learning. The structure already exists in Telecom; the industry has shared towers, cores, and standards for decades. The next telecom standard should not be about air interfaces, but how networks learn. Owning the cognition layer determines who controls optimisation, automation, and cost efficiency in the next cycle. Three operators have proved it’s technically possible. The question now is whether the rest will keep renting cognition or start building it. https://lnkd.in/gMJsk2Cy

At MWC Barcelona this year, we launched the GSMA Open-Telco LLM Benchmarks to unite a community tackling the unique challenges of telecom AI. The first results were clear: out-of-the-box AI models simply aren’t fit for telco-specific needs. Now, with version 2.0, this effort has evolved into a thriving, open-source collaboration. The findings point to a hybrid architecture as the most effective path forward - combining the broad reasoning of foundation models with the precision of specialised components. In addition to providing clear direction for AI in telecom, what’s really exciting is the unprecedented level of industry collaboration. Operators including AT&T, China Telecom Global, Deutsche Telekom, du, KDDI Corporation, KPN, Liberty Global, Orange, Telefónica, Turkcell, Swisscom, and Vodafone are joined by research and technology partners - Adaptive AI, Datumo, Huawei GTS, Hugging Face, The Linux Foundation, Khalifa University, NetoAI, Universitat Pompeu Fabra - Barcelona (UPF), The University of Texas at Dallas and Queen's University - to build a shared ecosystem for experimentation, validation, and learning. Read more in our latest blog: https://lnkd.in/eTDH5PBX

⚡ AI in Telecom: Beyond the Cost-Cutting Trap Let’s be honest- when most telecom execs talk about AI, the first slide is usually about cost savings: automate support, reduce truck rolls, optimize ops. Important? Absolutely. But if AI in telecom stops at cost-cutting… we’re missing another important play because you can only reduce the cost as much. 📡 The other opportunity lies in growth + customer value. ✨ Imagine AI that: - Predicts when customers are about to churn — and triggers personalized retention offers. - Designs dynamic, usage-based pricing models that adjust in real time. - Powers localized network slices for enterprises, hospitals, or smart cities (Naas). - Turns billions of IoT signals into new revenue streams. - Does Data Monetization & Partnerships This isn’t about trimming fat. It’s about reshaping the business model. The cost-cutting narrative makes AI sound like an efficiency tool. But AI can be the engine for innovation, differentiation, and growth in telecom if we identify the right use cases and work on them one by one. 💡My takeaway: AI will deliver savings, yes. But the winners will be those who go beyond efficiency and use AI to reimagine products, services, and customer relationships. 👉 Question: Is your AI strategy framed as a cost center… or a growth driver?

We’ve seen this story before. 2015: “Self-driving cars in 2 years” 2016: “Radiologists obsolete in 5 years” 2024: Reality check. Now fast forward to Telco & Networking: “AI will run networks autonomously.” “Zero-touch operations will eliminate NOCs.” “Self-healing networks will remove human intervention.” Let’s be precise. In telecom environments—especially large-scale DC fabrics, MPLS cores, and multi-vendor ecosystems—complexity doesn’t disappear. It compounds. AI will transform operations—but not by replacing engineers. It will augment decision-making, reduce MTTR, and surface anomalies faster than humans can detect. What actually works today: • AI-assisted fault correlation (noise reduction across millions of events) • Configuration drift detection and compliance enforcement • Predictive capacity and hardware failure insights • Intelligent automation pipelines (closed-loop, but supervised) What doesn’t (yet): • Fully autonomous networks without human guardrails • Black-box AI making production-impacting decisions • One-size-fits-all models across heterogeneous telco stacks The takeaway for Telco leaders: 👉 Don’t chase hype cycles—engineer for reality 👉 Focus on incremental AI adoption with measurable outcomes 👉 Build observability + data pipelines first, AI later 👉 Keep humans in the loop—especially for critical control planes AI in networking is not a replacement strategy. It’s a force multiplier for operational excellence. #Telco #Networking #AI #NetOps #Automation #NOC #Observability

🚀 The NVIDIA–Nokia Partnership: Thinking Beyond Telco While most headlines shout about 6G, Jensen Huang’s move with NVIDIA and Nokia is about something much bigger. This isn’t about radios — it’s about geography. The telecom network is the only industrial grid on the planet with power, fibre, and cooling everywhere. That makes it the perfect foundation for the next economy — a physical economy powered by robots, humanoids, and autonomous vehicles. To make that work, AI will need 10x more compute and 10x more reach than today’s cloud. And only telecom offers that reach. With NVIDIA’s new ARC platform — powered by Grace CPUs, Blackwell GPUs, and Spectrum X networking — every radio site can become a compact data centre. First, to run radio access. Next, to handle edge inference for machines operating in the real world. This is intelligence placement — compute positioned where the physical world lives and acts. The network that once carried voice will soon carry cognition. 🔹For Nokia, it’s a return to relevance. 🔹For operators, it’s a chance to monetise intelligent infrastructure. 🔹For NVIDIA, it’s about controlling the geography of AI. We’re already seeing glimpses of this future. T-Mobile’s AI-RAN Innovation Center with NVIDIA is exploring how RAN sites can double as compute nodes — running traditional network workloads and AI inference at the edge. But there’s a fascinating regulatory angle here too. Will authorities be comfortable with AI companies accessing or influencing RAN hardware — one of the most sensitive layers of telecom infrastructure? It’s a bold move that blurs the line between network operator and AI platform. One thing’s for sure: Jensen Huang is in the Telco room — and he’s thinking bigger than anyone else. #Telecom #NVIDIA #Nokia #EdgeComputing #AI #AutonomousSystems #6G #FutureOfAI #TelcoInnovation #ArtificialIntelligence #DigitalInfrastructure #NetworkTransformation

For decades, CSPs poured billions into 4G, 5G, spectrum, and radio networks — yet much of the digital value was captured above the network. Why? Because the value is not in infrastructure alone but also in the intelligence layer. In 2024, the top 100 operators generated $1.75T in revenue — but spent $1.38T in OPEX. That’s a massive opportunity to unlock profits. And it won’t be solved by more spectrum or faster radios. It will be solved by AI, automation, and a rethink of OSS/BSS and IT — from back-office systems into engines of growth, monetization, and customer experience. Legacy silos across billing, CRM, and network data are holding back innovation. Modern data platforms harnessing graph datasets and agentic AI will help change the trajectory — turning raw data into real-time intelligence that can act, orchestrate, and monetize. We’re already seeing it happen: > AI-driven anomaly detection and traffic prediction > Digital twins optimizing network energy use > Intent-based automation cutting order-to-cash cycles > GenAI agents accelerating catalog migration and product design Suppliers like Ericsson are embedding AI and automation across OSS/BSS to help CSPs reclaim control of key revenue levers. The winners will be those who shift investment toward the intelligence layer — building platforms that activate data, scale automation, and create new revenue streams. The question isn’t who builds the fastest network anymore. It’s who builds the smartest platform. For more on how this is being applied check out: https://lnkd.in/ehZhAk3h #Telecom #AI #Automation #OSS #BSS #AgenticAI #5G #NetworkAutomation #DigitalTransformation #AppledoreResearch #InnovatorsDilemma

Telecom has always evolved in waves. Each one reshaping not just networks, but the way leaders think. Before GSM, we operated siloed systems until GSM MoU asked a bold question: what if we aligned? That decision unlocked 3GPP, LTE surge, and ultimately 5G NR - transforming telecom from infrastructure into the digital backbone of economies. Now we’re at the next wave. Only this time, the catalyst is AI. And with AI the data is clear: 50%+ of CXOs already see noticeable gains in decision speed, foresight, and creativity through AI, and active use is expected to more than double within three years. At the same time, only 1% expect AI to make autonomous strategic decisions - a reminder that leadership judgment still sets the direction. Three things matter now: 1. AI doesn’t replace decisions, it improves the right ones. From capacity planning to churn prediction and service design, AI elevates decision quality while leaving human‑led calls where stakes are reputational. 2. Human AI chemistry becomes a leadership differentiator. Just as GSM succeeded through alignment, today’s leaders must learn to think with AI. Tools will standardize; judgment and collaboration won’t. 3. Governance is the accelerator. With 71% of CXOs citing legal and security risks, trust, explainability, and responsible data practices are what enable AI to scale safely and confidently. We’ve been here before - GSM, 3GPP, LTE, 5G, NTN. Each leap required leaders to rethink how decisions get made. AI is simply the next leap. And the leaders who lean in now will shape what comes next. https://lnkd.in/gKBNB_cu