I’m building a platform where small and mid-sized companies can host the internal apps they build themselves. In those conversations one question comes up early:
“Could that also run on our infrastructure?”
For a long time I treated it as yes or no, and I disliked both answers. “Yes” is cheap to say and expensive to mean. “No” ends a conversation with someone who was actually interested.
Then I read a piece by Omnistrate on bring-your-own-cloud. What stuck with me was the framing. They split the question into rungs. Roughly:
- We host it. Our servers, our account, our cloud bill.
- Your cloud account. You open the account and get the invoice, we deploy into it and operate it for you.
- Your network. It runs inside your existing VPC, behind your firewalls, your DNS, your routing.
- Your Kubernetes. You run the cluster, we ship a Helm chart that has to satisfy your policies.
- Air-gapped. No internet at all. Signed artifacts, moved in by hand.
Every step up puts you in a different company: different security review, different support burden, different sales cycle.
What made it click: nobody who asks that question means all five. They mean one of them, and for my buyers it’s almost always rung 2. When a 40-person Handwerksbetrieb or agency asks whether it can run on their infrastructure, they mean the data stays in Germany, the infrastructure belongs to them, and the cloud bill has their name on it. That’s the whole requirement. They have no service mesh and no platform team. Rung 4 is science fiction for them.
So my answer is one sentence now: “Rung 2, yes. Here’s where my no starts.”
Saying that out loud does something. Naming the limit sounds like you’ve done this before. “Sure, we can do that” sounds eager, and everyone who has ever bought software has been burned by eager. It also gives the other side something to push against: if they genuinely need rung 3, we both find out in minute five instead of month five.
The ladder isn’t a roadmap and I have no intention of climbing it. What it gives me is vocabulary, a way to answer in one sentence a question I used to hedge on, without over-promising and without closing the door. I’ve been the customer on most of these rungs myself, from shared PHP hosting in the late 90s to root servers to whatever we call today’s stack, and the honest reason I care about answering this well is that I’ve been on the receiving end of a vague yes.
If you build or sell software that has to end up in someone else’s infrastructure, decide which rung you’re on before a customer decides for you. The rung you choose is a business model. The rung you drift into is an accident.