NVIDIA and Major Telecoms Launch Distributed AI Grids at Network Edge

NVIDIA

Mar 20, 2026 · Updated Apr 25, 2026

NVIDIA partnered with AT&T, T-Mobile, Comcast, Spectrum, Akamai, and Indosat at GTC 2026 to build AI grids — geographically distributed inference platforms on telecom networks. Moving AI inference to the edge cuts latency and cost-per-token for real-time AI applications.

NVIDIA announced AI grids — geographically distributed AI infrastructure built on telecom networks — at GTC 2026. AT&T, T-Mobile, Comcast, Spectrum, Akamai, and Indosat Ooredoo Hutchison are partnering with NVIDIA to turn existing edge sites into AI inference platforms. Akamai is expanding across more than 4,400 edge locations with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, while Spectrum's deployment spans over 1,000 edge data centers within 10 milliseconds of 500 million devices.

This signals a structural shift in how AI is delivered — from centralized cloud to distributed intelligence at the network edge. Telecoms hold roughly 100,000 distributed data centers worldwide with over 100 gigawatts of potential AI capacity, turning existing network real-estate into a compute layer.

Explore the NVIDIA AI Grid Reference Design to evaluate deploying real-time AI applications — conversational agents, vision AI, or interactive media — on telecom edge infrastructure instead of centralized cloud.

View the full update on blogs.nvidia.com

NVIDIA AI

@NVIDIAAIMar 17

NVIDIA and telecom leaders are building AI grids to optimize inference on distributed networks. 🌐 @ATT, @TMobile, @Comcast, and @GetSpectrum are building these grids using NVIDIA AI infrastructure. @PersonalAI, Linker Vision, @ServeRobotics, and @DecartAI are deploying real-time AI applications across the grid. Read the #GTC26 announcement: https://t.co/V84GIwYIwe

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from NVIDIA →

Keep reading

NVIDIA transforms telecom networks into distributed AI grids for real-time edge agents

NVIDIA launched the Metropolis VSS 3 Blueprint, a reference architecture that allows AI agents to run directly on 5G telecom infrastructure. This shift enables continuous real-time video analysis and infrastructure monitoring at the edge, bypassing the latency and cost of cloud-based processing.