I spent a good amount of last year talking about the reasons why companies are moving away from hyperscale cloud, and what it takes to do it properly. Since then, I've had countless conversations with CEOs, CTOs and developers all looking to claw back control of their infrastructure.
Fast forward to 2025 and the trend hasn't only continued, but accelerated. Cloud repatriation is no longer a fringe idea scribbled on brainstorm whiteboards. It's now on the boardroom agenda, and companies are waking up to the fact that reducing their reliance on hyperscale cloud infrastructure can mean more growth in their business.
If the last few years were about cloud adoption, this year is about cloud maturity; understanding where hyperscale resource is needed, and optimizing infrastructure for where it isn't. Interestingly, I've picked up on three key trends this year that continue to take companies down the road of repatriation as that maturity sets in.
Let’s start with the most cited reason for reverse cloud migration – cost. Cloud spend continues to be a major pain point, not just for startups who took advantage of free credits to get their business off the ground, but also for well-established enterprises.
In fact, 17% of the 759 cloud decision-makers surveyed for Flexera’s 2025 State of the Cloud Report said they exceeded their cloud budgets.
What’s changing in 2025 is how visible that pain is becoming. CTO of 37Signals David Heinemeier Hansson made waves in the infrastructure space when he announced the company will be deleting their “entire AWS account this summer when the data is out”, and that “the savings will rack up pretty quick from there.”
For years, David has been vocal about the savings you can unlock by moving a part of your infrastructure off the cloud, never mind going all-out like 37Signals. And it’s no surprise either: 84% of respondents who took part in the Flexera 2025 report said managing costs is the biggest challenge they face.
This is generally down to three main reasons:
At this point, it feels like myself, David, and many others are preaching to the choir. So why isn’t the answer as simple as just leaving?
If cost is the wakeup call that hyperscale might not be the way forward, vendor lock-in is the reason businesses are stuck paying those prices for so long. The dirty secret of hyperscale cloud is that once you’re in, getting out is incredibly hard.
Picture this: you are the CEO of a start-up and you’re itching to get your products and services out there. You look at the big names of AWS, Azure, and Google Cloud, and see that not only do they offer free credits to get you up and running at a discount, but they also have out-of-the-box services that require no technical expertise to manage. On the surface it sounds like a win-win – who wouldn’t want to join a hyperscaler?
The catch comes when the free credits end. Suddenly the prices inflate, and businesses are left questioning how these costs are calculated. But pulling the plug isn’t as straightforward as it should be.
Those ready-made products that are seen as a convenient solution are purposely made to keep you in the hyperscale ecosystem. Services such as relational database as-a-service and firewall solutions are often unique to one provider’s environment, and require re-engineering and significant internal training just to get back to a vendor-neutral footing.
So, when a CTO approaches us for a reverse cloud migration, the first thing I always ask is: ‘what products are you using?’
Why? Because it is the products alone that turn what could be a simple migration into something incredibly complex – so much so, even the best technical teams are reluctant to undertake the task.
For any business tempted by offers from the likes of AWS and Azure, the first step before onboarding needs to be developing an exit plan. If you don’t have one, it’s time to create a cloud exit strategy.
Despite the challenges of vendor lock-in, 2025 is the year many companies are moving beyond just talking about cloud exit strategies – they’re actively implementing them in pursuit of a hybrid infrastructure. Often, that means a mix of cloud and bare metal to strike the right balance. Flexera’s 2025 report finds:
“This new normal is also reflected when asking about which clouds organizations choose to use. Looking at multi-cloud strategies, 70% of respondents embrace hybrid cloud strategies, using at least one public and one private cloud, while the remaining 30% use only public cloud(s) or private cloud(s).”
Hyperscale cloud has its place, especially within the industries of gaming and streaming where high infrastructure scaling needs are critical due to unpredictable demand levels. But there are ways to guarantee your demand is met, without solely relying on cloud.
When our Gaming Specialist Charlie Bowers asked Rob Schoeppe, CCO of AccelByte, for his tips on optimizing server costs, his answer was simple:
“For the usage you’re sure to have, run that on bare metal. But for any spike in usage - maybe there’s an event or a promotion that’s happening - you can flex into the cloud, whether that be AWS or Azure.”
At the same time, hyperscale growth hasn’t stopped. What we’re instead seeing is a more self-aware, intentional approach where businesses are placing steady workloads on bare metal or private infrastructure, and using cloud selectively, just as Rob illustrates.
It’s certainly a more thoughtful way to scale, prioritizing optimization of both resource and cost. I often say to customers: “don’t be 100% on us. But don’t be 100% on any one platform either.”
Hyperscale cloud isn’t going anywhere. The likes of AWS and Azure are still growing with increased spend year on year. But, almost paradoxically, infrastructure alternatives such as bare metal server hosting are gaining momentum too.
According to VMware’s Private Cloud Outlook study, 70% of respondents are actively considering cloud repatriation, with 35% already repatriating some workloads.
I believe these three trends - cost, vendor lock-in, and the rise of hybrid infrastructure - are forcing tech leaders to take a closer look at what truly belongs in the cloud and what doesn’t.
If any one of these cloud repatriation trends resonate with you and you’re wondering when the right time is to put a plan in action, I would love to have a chat. You can find me directly on LinkedIn, or can fill out our contact form to speak to our experts here.
With previous roles at Multiplay and PEER 1 Hosting, Isaac’s career has spanned all aspects of the global hosting industry including the development, procurement and selling of bare metal, colocation and cloud infrastructure.