The workload profile where repatriation genuinely pays off
Steady, predictable, compute- or storage-heavy workloads with low variance and a multi-year time horizon are where owned infrastructure's economics beat elastic cloud pricing — because you stop paying an elasticity premium for flexibility you're not actually using.
Bursty, unpredictable, or rapidly scaling workloads are the opposite case, and moving those off cloud on the strength of someone else's repatriation story is a straightforward modelling error.
What the honest cost model actually includes
Owned infrastructure's true cost includes depreciation, facility, networking, and the operational staff to run it reliably — not just the hardware purchase price compared against a cloud bill. Skipping any of those in the comparison is how repatriation announcements end up looking better on paper than they turn out to be in year two.
Repatriation is a workload-level decision wearing a company-level headline. Model the workload, not the headline.MNB Cloud Practice
Key actions
- Model workload variance before comparing owned versus cloud economics
- Include depreciation, facility, and operational staffing in the true cost
- Reserve repatriation for steady, predictable, multi-year workloads
- Avoid generalising one workload's repatriation into a company-wide reversal