NVIDIA and telecom operators deploy distributed AI grids for ultra-fast services

Telcos are transforming their networks into AI grids, powered by NVIDIA infrastructure. These distributed platforms bring AI inference closer to users and devices, enabling real-time applications from smart cities to cloud gaming while reducing latency and cost.

Nathaniel Alexander Abaño March 21, 2026

Telecommunications operators are transforming their networks into “AI grids,” using NVIDIA’s AI infrastructure to run real-time applications closer to users, devices, and data, the chipmaker said Tuesday.

The move marks a major shift in how artificial intelligence (AI) is deployed, putting telcos at the center of AI delivery rather than simply carrying traffic.

At NVIDIA’s GTC 2026, leading US and Asian operators unveiled plans to leverage their distributed network infrastructure, including edge data centers, mobile switching offices, and regional hubs, to create interconnected AI grids. These grids repurpose existing network assets and spare power, turning tens of thousands of sites into geographically distributed computing platforms capable of running AI inference efficiently and at scale.

Distributed cloud expansion

Different operators are taking varied approaches. Many are starting with existing wired edge sites, while others are integrating AI directly into their radio access networks (AI-RAN), enabling simultaneous connectivity and AI workloads. Collectively, telcos and distributed cloud providers operate some of the largest infrastructure footprints in the world, with the capacity to offer more than 100 gigawatts of AI compute over time.

AT&T, for example, is building an AI grid for IoT in partnership with NVIDIA and Cisco, running inference closer to where data is generated. Applications include real-time public-safety monitoring with Linker Vision, allowing faster detection, alerts, and secure data handling at the network edge. “Scaling AI services that are secure and accessible for enterprises and developers is a core pillar of our IoT strategy,” said Shawn Hakl, senior VP of product at AT&T Business.

Comcast is converting its broadband footprint into an AI grid for hyper-personalized media and cloud gaming, while Spectrum is deploying edge GPUs across its fiber network to render high-resolution graphics for media production. T‑Mobile and Indosat Ooredoo Hutchison are piloting AI-RAN and localized AI services, supporting smart-city, retail, and industrial applications. Akamai is expanding its globally distributed AI inference cloud across thousands of edge sites to power real-time experiences for gaming, media, finance, and retail.

AI grid ecosystems

Real-time AI applications are already benefiting. Personal AI is running conversational agents with sub-500ms latency, Linker Vision is enabling cities to detect accidents and unsafe crowd behavior faster than ever, and Decart is delivering interactive live video with sub-12ms latency at peak viewership.

The NVIDIA AI Grid Reference Design and a growing partner ecosystem — including Cisco, HPE, Armada, and Spectro Cloud — provide the full stack for deploying, orchestrating, and monetizing distributed AI workloads. Masum Mir, Cisco’s SVP of provider mobility, said, “Physical AI is accelerating the shift from centralized intelligence to distributed decision making at the network edge, enabling operators to participate fully in the AI value chain.”