When the hyperscaler is the wrong answer.
We have implemented and run a full OpenStack cloud as a platform service. For workloads where cost, residency or control decide the architecture, a private cloud stops being nostalgia and starts being arithmetic.
Talk to a tech lead ↗Four situations where private infrastructure wins on the numbers.
OpenStack, run as a platform rather than a project.
Implementation
Compute, storage, networking and identity stood up as a coherent platform — not a pile of components a team is left to integrate.
Platform engineering
The layer your developers actually touch: provisioning, pipelines, observability and the guardrails that stop a private cloud becoming a pet.
Operations
Upgrades, capacity and incident response on infrastructure that has to stay up. Boring infrastructure wins, and staying boring is work.
Hybrid, honestly
Most answers are mixed. We will tell you which workloads should stay on a hyperscaler rather than move everything to justify the engagement.
Private cloud, asked properly.
Is a private cloud cheaper than a hyperscaler?
For steady, predictable, egress-heavy workloads it can be. For spiky or early-stage systems it usually is not. The honest answer depends on your utilisation curve, which is why the first conversation is an architecture conversation.
Does this mean running our own hardware?
Not necessarily. Private cloud describes the control model, not the building. Dedicated or colocated infrastructure under your governance covers most sovereignty requirements without anyone buying a data centre.
How does this help with data residency?
You get to say exactly where data lives and who can reach it, and demonstrate it. We describe practices rather than sell certifications — the architecture is what you can point at in an audit.
What if we only want part of this?
That is the normal case. Repatriating one workload is a smaller, more reversible decision than a platform strategy, and it is a better way to find out whether the arithmetic holds.