Cloud Repatriation: Why Companies are Bringing Data Home

Right now, something counterintuitive is unfolding across enterprise IT departments. Companies that spent the last half-decade migrating to the cloud are now pulling workloads back on-premise. Not because the cloud failed, but because it succeeded too well at something nobody anticipated: becoming extraordinarily expensive at scale. 

If you’re an IT director watching your cloud bill climb month after month, you’re part of a growing conversation. Major enterprises like Dropbox and 37signals are publicly documenting their cloud exits and sharing the financial impact. This shift isn’t about rejecting the cloud — it’s about finally getting strategic after learning what “pay as you grow” really costs when you’ve actually grown. 

Here’s what’s driving this moment: cloud costs for mature workloads are outpacing IT budgets, data sovereignty laws are dictating where data must live, and the technical debt of cloud-native architecture is coming due, unwinding it costs more than anyone planned for. 

The cost problem is catching everyone off guard 

Companies migrated, assuming cloud pricing would stay predictable. It didn’t. 

Egress fees, charges for moving data out of cloud infrastructure, are becoming budget killers. Organisations that adopted aggressive cloud deployment policies often find themselves overpaying for configurations that no longer make economic sense, leading to “bill shock” among stakeholders. 

The pain point isn’t the initial migration, it’s the steady-state costs that accumulate once you’re fully committed. When workloads run continuously in cloud environments, companies discover they’re paying far more than equivalent on-premise infrastructure would cost. 

Cloud-native tooling also creates vendor lock-in that’s expensive to escape. IT leaders are realising this isn’t a temporary pricing spike, it’s the business model working exactly as designed. 

When sovereignty becomes non-negotiable 

Cloud repatriation isn’t purely economic. Regulatory pressure is forcing hands right now. 

The EU’s Digital Operational Resilience Act (DORA) entered into application on January 17, 2025, requiring financial institutions to strengthen digital resilience and manage ICT risk more directly. France’s Doctrine Cloud mandates that government data stays in French-controlled facilities, while Germany’s GAIA-X initiative aims to reduce dependence on US hyperscalers. 

This is fundamentally different from earlier compliance regimes. GDPR allowed workarounds through legal agreements. Modern data sovereignty laws demand physical infrastructure within borders, under local jurisdiction. For organisations operating in these regions, a cloud repatriation strategy is becoming mandatory, not optional. 

Who’s actually moving what? 

Not everything comes back. That’s the nuance missing from most coverage. 

Organisations are adopting hybrid approaches, most repatriating only select parts of their workloads back to on-prem or private environments, rather than doing a full exit. Financial services and healthcare are leading due to scale and regulation. Tech companies with strong infrastructure teams are following close behind. 

A pattern is emerging across industries: 

Barclays’ CIO Survey at the end of 2024 showed that almost all respondents planned to move some public-cloud workloads back to private cloud or on-premises, the highest on record. This trend is accelerating. 

What started as a cost-optimisation discussion has become an urgent strategic pivot as IT leaders realise cloud expenses are consuming unsustainable portions of technology budgets. 

Real-world repatriation success stories 

The numbers are striking. 

• Dropbox saved nearly $75 million over two years by bringing big-data workloads home to colocation facilities, cutting AWS usage to 10%. 

• Software company 37signals saved $2 million in 2024 by reducing cloud reliance and projects $10 million in savings over five years. 

For predictable workloads, owning the infrastructure simply makes more financial sense. 

The hidden complexity tax 

Reversing cloud migration isn’t just “move servers back.” It’s architectural surgery. Applications built cloud-native assume infinite scaling, managed services, and APIs that don’t exist on-premise.

Rebuilding applications designed around AWS Lambda or Azure Functions for on-prem deployment isn’t trivial, you’re essentially redesigning how the application operates. 

A talent gap compounds this. Many engineers hired during the cloud era haven’t worked with physical infrastructure before. Organisations often need to rebuild institutional knowledge. And vendor contracts make leaving painful — multi-year agreements create penalties for early exits that can run into millions. 

What’s actually working right now? 

Organisations seeing success with cloud repatriation are using workload economics as their north star. Three approaches are delivering results: 

Profile before you move: Identify the top workloads by volume, cost, and sovereignty requirements. Moving even a handful can deliver outsized savings. 

Negotiate hard first: IDC’s Cloud Pulse survey shows nearly half of cloud buyers spent more than expected. Reserved instances and contract renegotiations can dramatically change the economics — sometimes without moving anything. 

Consider colocation as a middle ground: Colocation facilities provide the control of on-prem without full data-centre ownership. Hyperconverged infrastructure and Kubernetes on bare metal bring cloud-style agility to private environments. 

You’re not returning to legacy architecture; you’re building cloud privately. 

The market reality 

This might seem to contradict the repatriation narrative, but only about 21% of workloads and data have been repatriated so far. Meanwhile, cloud adoption continues to grow in parallel. IDC’s June 2024 report found 80% of organisations expect some level of repatriation within the next year. The shift is not binary — it’s about workload placement strategy. 

Distilled 

Cloud repatriation isn’t about going backward. It’s about finally getting strategic instead of following the herd. Variable workloads, development environments, and burst capacity? Cloud makes sense. 
Predictable databases, AI training clusters, and sovereignty-sensitive data? On-premise wins. 

The question isn’t whether you’ll rethink your cloud strategy. It’s whether you do it proactively — before your CFO mandates it when the next budget lands on their desk.