NVIDIA and Telecom Leaders Build AI Grids to Optimize Distributed Network Inference

As AI-native applications scale, telecom networks are emerging as a key battleground for AI distribution. Major telecom operators are building geographically distributed AI grids to leverage existing network infrastructure to deliver and monetize AI services across the edge.

Operating methods vary. Some leverage existing wired edge sites, while others build edge inference platforms through AI-RAN technology that integrates AI into radio access networks. The core idea is to realize low-latency, low-cost inference near users by utilizing approximately 100,000 distributed network data centers worldwide and over 100GW of spare power.

Examples from major global operators are as follows:

• AT&T: Partnering with Cisco and NVIDIA to build an IoT-dedicated AI grid, supporting real-time public safety and mission-critical applications.
• Comcast: Leveraging low-latency broadband infrastructure to secure responsiveness and economics for conversational agents and NVIDIA GeForce NOW.
• Spectrum: Supporting high-resolution graphics rendering through over 1,000 edge data centers.
• Akamai: Deploying NVIDIA RTX PRO 6000 Blackwell GPUs across over 4,400 edge locations to provide intelligent orchestration.
• Indosat Ooredoo Hutchison: Connecting sovereign AI factories with AI-RAN sites for local innovation in Indonesia.