We (the company I work for) are looking at it from an IoT perspective. There’s a lot of compute that the devices we deploy just can’t handle, but we need the low latency and regional/geographic context to process appropriately. We’ve considered using some of the CDNs and cloud compute at the edge but even then that wasn’t quite capable enough for us, so we have deployed our own compute units to our own edge locations to run our workloads (mostly computer vision and aggregations). There’s some roll up data that gets reported upstream but most of it goes back to the devices for use there.

(I’m being deliberately vague here for business reasons, please forgive me.)