What happens if you repatriate workloads just to escape a bill?

Azati warns that repatriating purely to escape a cost surprise just recreates the same bill on-prem with worse elasticity. The underlying architecture problem follows the workload wherever it's hosted.

What discipline does Azati recommend instead of blanket repatriation?

Azati recommends honest workload placement and FinOps discipline, moving what genuinely benefits from repatriation and leaving what doesn't. This workload-by-workload approach avoids the all-or-nothing thinking behind most repatriation headlines.

Is the 2026 Cloud Repatriation Wave a Real Correction – or Just FinOps Blah-Blah?

Back to quick insights

August 18, 2026

Is the 2026 Cloud Repatriation Wave a Real Correction – or Just FinOps Blah-Blah?

Maksym Kramarenko

Digital Transformation Advisor, Azati

Quick answer: Ask a room of engineers for the big infrastructure story of 2026 and a surprising number say the same odd thing: people leaving the cloud. The stat everyone quotes is 86% of CIOs planning to move some workloads back off public cloud (Barclays, end of 2024), the highest ever recorded. Dig in and it gets murkier: only about 8% plan a real exit, and public cloud spend is still climbing past $700B. So the truth is boring and specific: it’s a real correction for some workloads, and theater for the rest.

One view: a genuine architectural correction

Plenty of workloads were lifted-and-shifted into the cloud that never belonged there, and steady-state compute can be dramatically cheaper on private infrastructure.

The other view: it’s a statistical illusion

The 86% measures movement, not direction – moving one compliance-bound database and exiting the cloud entirely count the same in that survey. Meanwhile the spend line goes up, not down.

Our read

Most "repatriation" is teams blaming the cloud for a bill their own architecture ran up. The cloud didn’t overspend – an unexamined design did. The fix isn’t a mass migration; it’s honest workload placement and FinOps discipline, moving what genuinely benefits and leaving what doesn’t. If you’re repatriating to escape a bill, you’ll just recreate the same bill on-prem with worse elasticity.

FAQ

Is everyone really leaving the public cloud?

No. About 86% of CIOs plan to move some workloads back, but only ~8% plan a full exit, and total cloud spend keeps rising past $700B.

When does cloud repatriation actually make sense?

For predictable, steady-state workloads with stable demand – not as a blanket reaction to a cost surprise. Start with a workload-by-workload assessment.

Not sure which workloads belong where?

Let’s do the architect’s read, not the vendor’s

Adapted from a LinkedIn post by Maksym Kramarenko, Digital Transformation Advisor at Azati – read the original. Sources: Barclays CIO Survey (Q4 2024); IDC; Gartner.