Comment by Dutchhack

To be clear, I’m not arguing against hyperscalers they’re obviously extremely good at what they do. The question is whether all device-adjacent workloads should default to them, or whether there’s value in introducing an autonomous first hop that can enforce locality, policy, and failure containment before anything goes upstream.