Edge AI inference compute to piggyback on US telecom infra

Big, centralized data center projects in the US are running into delays. New construction methods, tighter standards, a lack of skilled workers and materials, and slow hookups to power and water are all pushing timelines out.

To work around this, some AI workloads are shifting closer to end users, riding on top of existing US telecom infrastructure instead of waiting for new hyperscale capacity. By placing local edge AI inference compute within or alongside telecom networks, providers can make use of already-built sites, power, and connectivity to serve latency-sensitive applications, even as large data center builds stall.