Most hosts our size rent capacity from someone larger and resell it. We own the machines, the racks they sit in and the switches above them. That is a more expensive way to start, so it is worth being specific about what it actually changes for a customer.
Hardware failures get fixed by us
When a disk fails in a machine we own, an engineer walks to the cabinet with a replacement. When a disk fails in rented capacity, a ticket is raised with the landlord and everyone waits. The mean time to repair is the clearest difference between the two models and it is not close.
We can say yes to unusual requests
A specific NIC, a second uplink into a different upstream, an out-of-band serial console, a locked cabinet with your own padlock — these are all decisions we can simply make. Under a rental agreement they are decisions somebody else makes on their own timeline, if at all.
The network is ours to change
We run our own ASN and our own peering. If a transit provider degrades we can shift traffic in minutes rather than opening a ticket about it. Route changes, new peering sessions and BGP communities for customers are all things we control directly.
Where it does not help
Owning the facility does nothing for you if your workload needs to be in Frankfurt and we are in Houston. It does not make a single machine more reliable than the sum of its parts — a machine with one power supply is a machine with one power supply, whoever owns the rack. And it does not make us cheaper than a hyperscaler at hyperscaler volume.
If your requirement is global edge presence, we are the wrong answer and we will tell you so. If your requirement is predictable hardware in one place, run by people who can physically touch it, that is the case for owning it.
Runs the edge and the maintenance calendar in HOU1. Writes the postmortems, including the ones that are not flattering.