Are Businesses Leaving the Cloud?

By Cameron Daniel, Chief Technology Officer

Learn the truth about cloud repatriation, the motivations behind it, and whether it’s really happening as much as you think.

For years, the cloud has been the default solution for businesses wanting speed of deployment with quick and easy scalability. And while the cloud promises endless resources at your fingertips, a lot of network teams are having the conversation about whether to pull their workloads back out of the public cloud and run them on their own hardware or private cloud again. You might hear this trend referred to as cloud repatriation.

But why are some businesses considering rebuilding on-premises? Are they actually going through with it? And do you have to leave the cloud to get a better network?

So, are businesses leaving the cloud?

When I speak to customers about their motivations for wanting to move workloads out of the public cloud, two reasons come up again and again.

First, the cost. The cloud isn’t cheap, and for many, the bills keep climbing as big data workloads increase. Meanwhile, budgets are pinched.

Second, complexity. Dev teams picking “the right tool for the job” makes sense in theory, but in practice it’s creating sprawl. Before long, the network environment can become so tangled that it’s hard to keep track of what’s operating where.

The cloud has always been a double-edged sword. It can give you speed and flexibility, but the trade-off is losing some control and predictability. But here’s the catch – while plenty of teams are talking about leaving the cloud, very few of them are going through with it. Why? Because leaving the cloud is far easier said than done.

How do you replace the conveniences that come with using a global cloud provider that gives you so many ways to deploy and orchestrate your workloads? Where do you now put those workloads, and how do you set them up so they can be accessed and used in the way you need them?

Plus, if you intend to start operating your business on your own infrastructure, you’d need to consider:

• Upfront costs: Unless you deploy Compute as a Service from providers like Latitude.sh, hardware requires capital investment that the cloud doesn’t.
• Maintenance: You’re responsible for patching, monitoring, and upkeep.
• Scaling: Growing quickly can be challenging without the elasticity of the cloud.
• Reach: Establishing a presence in new regions takes more time and effort compared to spinning up instances in a cloud region.

Cloud repatriation will not only disrupt your operations, it will also be a costly and complex undertaking of its own – one that isn’t necessarily worth it.

So I don’t think we’ll see many companies actually leaving the cloud. Instead, what I am seeing is more IT teams taking a step back and asking where each of their workloads should really live.

The resulting outcome could be a mix of bare metal, compute, hybrid cloud, and multicloud adoption to get best-of-breed features across a more diverse setup.

Recent moves in the industry—including Megaport’s acquisition of Latitude.sh to bring compute and connectivity closer together—reflect this shift. Teams want more control without losing the elasticity they’ve grown used to in the cloud.

Should you adopt bare metal?

A bare metal server is simply a physical machine dedicated to a single customer. No sharing resources with anyone else and no virtualization layers in between – you get the hardware to yourself.

Compared to the cloud, bare metal puts your systems back under your control as you install your own operating system and run your applications directly on the server. Compared to virtualized cloud machines, bare metal provides better performance and more predictable costs in many use cases.

Bare metal still plays a key role in modern architectures, especially for predictable, performance-sensitive workloads.

But bare metal does come with trade-offs:

• Upfront investment: Traditional deployments require the buying and provisioning of hardware.
• Time to deploy: Getting physical servers ready can take weeks compared to minutes in the cloud.
• Scaling challenges: Growing capacity on demand isn’t straightforward unless you have a surplus of hardware.
• Operational overhead: Ongoing patching, monitoring, and upgrades fall to your team.
• Geographic reach: Expanding into new markets requires new data center builds or colocations.

While traditional bare metal deployments require owning and managing physical hardware, on-demand compute offerings change the model entirely. Platforms like Latitude.sh remove upfront hardware investment through on-demand billing, deliver servers in seconds instead of weeks, and provide access to dedicated hardware across more than 20 global locations. This lets teams use bare metal where it makes sense, without taking on the operational and financial friction that usually comes with it.

The role of compute

This is where compute comes in: flexible, on-demand processing power that lets you run workloads wherever they make the most sense, without tying you to a specific infrastructure model. Compute isn’t limited to virtual machines or public cloud platforms – it can include bare metal, VMs, container-ready environments, and even serverless services, depending on the workload.

With modern compute services, you provision exactly the resources you need (CPU, memory, storage, and operating system) through a portal or API, and scale capacity up or down up or down as requirements change. Instead of managing physical servers directly, the provider handles the underlying infrastructure, giving you predictable performance without worrying about racking hardware, replacements, or maintenance.

This model delivers the elasticity people associate with the cloud, while still allowing workloads to run on dedicated hardware when performance or isolation matters. Compute and bare metal work together by combining flexible, abstracted capacity with single-tenant servers for consistent performance.

Platforms like Latitude.sh take this a step further by delivering bare metal through a cloud-style provisioning model. With Megaport’s acquisition of Latitude, customers can deploy compute on demand and connect it directly into their private, hybrid, or multicloud architectures using the same Megaport network fabric.

Hybrid cloud for balance

Bare metal isn’t your only cloud diversification path – hybrid cloud can offer the best of both worlds.

Hybrid setups let you run stable, predictable workloads on your own hardware or compute, while still using the cloud for things like scaling, backups, or workloads that need global coverage. It’s a way to mix and match so you’re not locked into a single approach.

Benefits of a hybrid cloud setup include:

• Flexibility: Place each workload where it runs best.
• Cost balance: Keep predictable jobs off the cloud, but use it for elasticity when needed.
• Resilience: Spread workloads across environments to protect business continuity.

For a lot of organizations, hybrid has become the “sweet spot”.

Multicloud for provider flexibility

Hybrid isn’t the only way to diversify your infrastructure. When you spread workloads across different cloud and SaaS providers, you take advantage of each provider’s unique strengths and pricing models. Compute can also sit alongside these cloud services to boost reliability and performance.

Multicloud offers benefits like:

• Cost optimization: Shift workloads to the provider offering the most competitive pricing or billing model for that service.
• Service specialization: Leverage best-in-class features across providers, such as GPU acceleration, analytics tools, or advanced security options.
• Risk mitigation: Reduce reliance on a single cloud to improve resilience against outages and regional service disruptions.

For example, with a multicloud approach, you might choose one provider for cost-effective storage, another for AI/ML services, and a third for its global edge footprint. For organizations running diverse workloads, multicloud is a way for you to stay agile while getting the best possible value.

Of course, a poorly managed multicloud can increase complexity, so it’s important to manage the network layer to optimize these benefits (more on that shortly).

Which path to choose

My advice: Businesses shouldn’t be (and aren’t) pulling wholesale out of the cloud.

Instead, the smarter move to improve your costs and management is to diversify.

Stable, predictable workloads belong on compute or bare metal, where you get the best performance and cost control.

Elastic or bursty workloads are better suited to hybrid cloud, giving you the balance of local infrastructure with the scalability of the public cloud.

And for organizations running diverse workloads or wanting to avoid lock-in, multicloud is the way forward – use each provider for what they do best.

Have all of the above? Use all three approaches for an architecture that is cost-effective and agile.

Even if you use all of these approaches, you can still keep workload management simple with the right network setup.

The critical point: the network

Whichever path/s you choose, the network is your glue.

Using Network as a Service (NaaS) gives you the kind of control that’s often missing from cloud or on-prem setups alone, marrying your different environments together on a single, flexible network underlay.

When you use a NaaS provider like Megaport to connect your workloads, you can:

• Switch between setups easily: Move workloads between compute, bare metal, hybrid, and cloud in just a few clicks via a simple portal.
• Control costs: Use private connections to avoid expensive cloud egress charges.
• Boost performance: Private, on-demand links reduce latency and improve reliability.
• Stay agile: Spin up new connections on-demand instead of waiting weeks to deploy new connections.

With Latitude.sh now part of Megaport, customers can also run compute close to their cloud and network edge environments without giving up the operational simplicity they expect.

Cloud repatriation isn’t about abandoning the cloud altogether, but about making smarter choices for your business. Not every workload belongs in the cloud, and that’s okay. Some are better suited for bare metal. Many fit best in a hybrid setup.

The ability to stay flexible—adapting your infrastructure without getting boxed in by costs or complexity—is key. A virtualized network underlay keeps your options open, whichever direction you take.

As compute and connectivity converge, especially through platforms like Megaport with Latitude.sh, businesses have more flexibility than ever to choose the right home for every workload.

Build your perfect network setup.