- NVIDIA and carriers like AT&T, T-Mobile, and Comcast are building the “AI grid” — a distributed AI infrastructure running inference close to users to reduce latency and costs.
- The system leverages approximately 100,000 global network data centers, with the potential to provide over 100 gigawatts of new AI capacity.
- The AI grid shifts the role of telcos from data transmission to directly operating and monetizing AI services.
- AT&T deploys the AI grid for IoT with over 100 million connections, supporting real-time applications like public safety via Linker Vision.
- Comcast and Akamai optimize AI experiences such as conversational agents and cloud gaming with low latency and significantly reduced token costs.
- Spectrum is building a network of over 1,000 edge data centers, serving 500 million devices with latency under 10 milliseconds.
- Practical applications: Conversational AI (Personal AI), urban surveillance (Linker Vision), and personalized real-time video (Decart).
- Outstanding performance: Latency under 500 ms, over 50% reduction in cost per token, 10x faster accident detection, and 15x faster disaster response.
- Conclusion: NVIDIA and telecom carriers like AT&T, T-Mobile, and Comcast are building the “AI grid” — a distributed AI infrastructure running inference close to users to reduce latency and costs. The AI grid is reshaping the entire telecommunications industry by turning networks into distributed AI computing platforms. With 100,000 data centers and hundreds of gigawatts of potential, carriers are no longer just transmitting data but directly providing AI services. Thanks to ultra-low latency (under 10 ms) and costs reduced by over 50%, AI-native applications such as smart cities, real-time video, and conversational agents will explode, making telcos the new hub of the AI economy.
