More companies are moving some of their workloads off public cloud, and the interesting part is what they are moving and why. This is a sorting exercise, where each workload goes to the place that suits it, whether that is a major provider, private infrastructure, or a mix of both. The useful question for your own business is which of your workloads actually belong where they are today.
The shift is real, and it is happening in the open
Something has changed in how companies talk about the cloud. For a decade the default answer to where a workload should run was a major public provider, and questioning that was unusual. In 2026 the questioning is mainstream. A Barclays survey of enterprise technology leaders found that around 83 percent plan to move at least some workloads off public cloud towards private or on-premises infrastructure, and other 2026 surveys put the share considering it at record highs. Running across public cloud and a company’s own infrastructure at once, often called hybrid, is becoming the ordinary way to operate.
Why the maths changed for steady work
The reason is mostly about the shape of the work. Public cloud pricing rewards workloads that rise and fall, because you pay for what you use and drop the cost when demand drops. Workloads that run flat out around the clock are a different case, and there the pay-as-you-go premium adds up. Andreessen Horowitz described this as the cloud paradox, estimating that heavy public cloud use can weigh a software company’s gross margins down by half or more once it reaches real scale. The company behind Basecamp, 37signals, projected around ten million dollars in savings over five years after moving its own steady workloads off a major provider. Numbers like that are why the question now reaches the board.
So should you leave the cloud?
For most companies the honest answer is no, or at least not entirely. The cloud still earns its place for a lot: variable or hard-to-predict demand, early teams that need to move without buying hardware, and work that scales up and down with your users. The companies moving workloads back keep all of that where it is. What they move is the steady, heavily used work that runs cheaper on fixed capacity. Below a certain level of utilisation the cloud is the better deal, and above it the sums start to favour owning or renting capacity of your own.
How to think about your own setup
You can do a version of this analysis without a large project. A few things are worth looking at:
How steady each workload is. Something that runs at a constant load all day is a candidate to price elsewhere, while a service that spikes with traffic usually belongs where it can scale on demand.
How well used your committed capacity is. Machines sitting far below what you are paying for are charging you a premium for flexibility you are not using.
How portable the workload is. Something built deeply on one provider’s own services is harder to move, so the comparison becomes renewal leverage rather than a real switch.
Data gravity and compliance. Large volumes of data cost money and time to shift, and some data has to stay in a particular place for legal reasons.
The financial side of moving is getting easier too. From 12 January 2027 the EU Data Act removes cloud switching charges across the EU, and the major providers already dropped egress fees for departing customers in 2024. The cost that remains is the engineering work of re-architecting, shifting data, and testing. That part is real, so a move earns its place only when the running savings clear it comfortably.
The takeaway is calmer than the headlines suggest. The choice has always been a workload-by-workload one, and the companies getting it right are the ones asking, honestly, where each part of their setup belongs.
