Back

3 hyperscale cloud repatriation examples to rethink your infrastructure strategy

3 cloud repatriation examples to rethink your infrastructure strategy

Organizations have started to feel the pressure of mounting bills and rising complexities in hyperscale cloud environments. And for some, it’s reached a critical impasse in which maintaining a 100% hyperscale cloud infrastructure has become a significant risk to profitability and control.

Even AWS has come out and said that they’ve noticed more of their customers actively looking to reduce reliance on AWS infrastructure. In a hearing held on July 2nd 2024, AWS told the Competition and Markets Authority (CMA) that its hyperscale cloud business is facing competition. Specifically, AWS exampled customers moving back to on-premises IT solutions in a bid to manage internal finances and increase ownership over resources, data and security.

But it’s not just on-premises solutions that are seeing a resurgence. Shifting priorities when it comes to infrastructure are driving a wider evolution in the cloud computing landscape. It’s driven largely by new service models like bare metal cloud that offer a middle ground between hyperscale cloud solutions and on-prem deployments.

The bare metal cloud market is projected to grow by a compound annual growth rate of 23.9% between now and 2030. And with more hyperscale cloud alternatives available than ever before, the barriers to exit are quickly coming down.

In this blog we’ll explore some examples of companies that made the move away from hyperscale cloud to alternative infrastructure solutions.

Understanding cloud repatriation

Cloud repatriation means the moving of workloads off hyperscale cloud services and onto alternative solutions. It often involves a return to on-premises hardware but can also include private cloud solutions, colocation setups, bare metal server hosting, or a hybrid approach. 

3 hyperscale cloud repatriation success stories

cloud repatriation

FxGrow’s transition from hyperscale cloud to bare metal cloud

For online forex brokerage FxGrow, migrating infrastructure out of cloud was primarily driven by security, support and performance concerns. The company had been using a combination of hyperscale cloud hosting services and specialized forex hosting solutions, but were experiencing errors, disconnections and even, on one occasion, a security breach.

“We kept getting disconnections and our servers kept going down. It was causing us constant headaches with our clients,” said Mohammad Mazeh, Head of Brokerage and Platform Administration at FxGrow.

For a company operating in such a performance and security-sensitive industry, this couldn’t continue. So FxGrow decided to migrate some of its infrastructure to a bare metal cloud hosting solution to gain control over their technology and access specialized support. With single-tenanted dedicated servers the team can operate in a secure environment with much more control over their technology.

“Our infrastructure is separated from everyone else. Nobody can access our servers unless we make a mistake from our end… our main trading server IP address is hidden so no client can access it,” said Mazeh.

The migration has allowed the brokerage to protect its clients and improve customer trust. As a result, FxGrow now plans to migrate all its remaining infrastructure to a bare metal cloud hosting solution in the near future.

Dropbox’s journey to colocation

Dropbox was one of the first big companies to talk publicly about migrating away from hyperscale cloud infrastructure. The cloud storage and file sharing service started using AWS S3 in 2013 because of its large storage capacity. But by 2015 the company decided to move most of its data out of AWS and into a custom infrastructure solution in colocation. They still use AWS for some workloads but 90% now sits in privately owned hardware.

It was a monumental move for a company with 500 million customers and 500 petabytes of customer files already in AWS. But even though the decision came with risks, Dropbox needed more control over their technology. Especially in preparation for the 2016 launch of Magic Pocket, Dropbox’s own multi-exabyte infrastructure designed to manage large capacity growth at scale.

“For us it was about quality and control management,” Dan Williams, Head of Production at Dropbox told TechCrunch in the aftermath of the migration. “We know there are solid parties out there with high quality and performance, but we felt ours could be equal or even better because we know the system so well”.

And even though it wasn’t the primary motivation, the move also brought with it a significant boost to profit margins. In the first year alone, Dropbox was able to save $39.5 million - the result of a $92.5 million decline in third-party data center expenses and a $53 million increase for their proprietary data centers. Cumulative savings over a subsequent two years reduced operational expenses by $74.6 million.

Today, Dropbox continues to run its infrastructure on a hybrid model combining on-premises hardware and public cloud.

Dukaan chooses managed bare metal

It’s not just economies of scale that have businesses reconsidering their infrastructure choices. In 2023, Dukaan – an ecommerce startup that helps local store owners to sell their products online – also left the cloud.

Public cloud infrastructure was integral to Dukaan’s rapid initial growth. The team launched their first minimum viable product within a mere two days. But soon infrastructure costs started to consume the majority of Dukaan’s budget (despite numerous attempts to optimize spending).

Monthly cloud bills in the region of $90,000 were just not sustainable. But rather than passing the premium to their customers through increased service fees, Dukaan decided to reduce costs at the infrastructure level by moving to a managed bare metal solution.

The savings were considerable. Monthly costs decreased from $90,000 to $1,500. At the same time, the ongoing latency issues Dukaan were experiencing in public cloud reduced significantly and the team gained more control over their hardware and network configurations.

The project was a resounding success. In the aftermath of the migration Subhash Choudary, posted on X with the following words: “the best way to save money on the cloud is to remove it. Dukaan now runs almost entirely outside of cloud”.

Final thoughts

Whether in a bid to gain control over their technology or to support future profitability, the rate at which businesses are abandoning hyperscale cloud-first strategies is increasing. And as more companies speak out about their own hyperscale cloud repatriation success stories, it’s likely that we’ll see others start to follow suit.

It’s a massive departure from the hyperscale cloud-first strategies that have dominated since the meteoric rise of AWS in the early 2000s. But it’s not to say hyperscalers are becoming redundant. Even David Heinemeier Hansson, arguably the most critical figure in opposition to hyperscale cloud right now, still stands behind its merits in specific circumstances.

Instead, what these examples serve to do is inspire others to take a hard look at their own infrastructure and ask: is it serving us?

Related articles