Even in cloud repatriation there is no escaping hyperscalers 

The cloud repatriation argument is bubbling up again as more enterprises eye a return to on-prem deployments, and  – hyperscalers are already plotting how to keep their hooks in enterprises that decide to jump ship. 

“The [hyperscaler] roadmaps all point to a future world where you can still use let’s say Google Cloud management tools on your on-prem environment,” Hariprasad Pichai, a principal in the Telecommunications, Information, Media and Electronics practice at consulting firm Arthur D. Little, told Silverlinings. “So, it’s a way for them to keep the lock-in while allowing you to do on-prem.” 

He added, “Their game is to create usage. It doesn’t have to be in their own data centers.” 

Cloud repatriation refers to bringing enterprise workloads back to on-premises environments. Pichai (no, not related to that Pichai) said the idea has been floating around since at least 2016 when Dropbox decided to abandon Amazon Web Services (AWS) and return to using in-house infrastructure. 

However, the back-and-forth over whether repatriation is a good idea erupted again – and this time louder – in late 2022 when 37signals (the company behind Basecamp and email app Hey) announced it too was leaving the public cloud. 

In a much-discussed blog, 37signals CTO David Heinemeier Hansson explained its reasoning for repatriation goes something like this: despite running on Amazon and Google’s infrastructure for years and seeing “all the cloud has to offer,” the company never realized the promise of reduced complexity and cost. So, it concluded operating in the cloud is a “bad deal” and decided to leave.  

“We're paying over half a million dollars per year for database (RDS) and search (ES) services from Amazon,” he wrote about Hey’s cloud expenses. “Do you know how many insanely beefy servers you could purchase on a budget of half a million dollars per year?” 

Cloud repatriation push back 

Hansson’s blog sparked pushback from cloud advocates like TelcoDR CEO Danielle Royston. She contended in an interview with Silverlinings that those using the cloud aren’t just paying for servers, but also for the proprietary tools the different cloud giants provide, the salaries they pay their top-tier developer talent, the hardware upgrades they make available to cloud users and the built-in security they offer.  

For those who use the cloud to its full potential, she said, the cloud is “the gift that keeps on giving.” Not only that, but those looking to repatriate workloads will need to invest significant time and money to transition back and hire more staff to develop new applications and manage the on-prem servers, she added. 

However, Hansson told Silverlinings that “going to the cloud didn't reduce the developer or operator overhead” associated with its workloads. He added that 37signals won’t need to increase its headcount to run its own hardware.  

So, who’s right? Well, it seems the answer will vary by company and even by application. 

Pichai explained the cloud is the ideal environment for a small handful of workloads, namely “vanilla applications” which incorporate only standard rather than specialized features and “spikey applications” which need to scale on demand to accommodate irregular patterns of usage.  

Companies with applications outside those categories, which may have “mindlessly shifted to the cloud” and seen costs balloon out of control are the ones reconsidering their spending, Pichai said. 

David Linthicum, chief cloud strategy officer at Deloitte Consulting, told Silverlinings applications which don’t use specialized cloud-native services are good candidates for repatriation. “Since they are not coupled to a particular public cloud provider and can be run just as well on premises, they could be repatriated, in many cases, at a lower cost and thus higher value to the business,” he said.  

Repatriation factors

Pichai added there are a few other things that factor into calculations around repatriation. 

First, he said there’s the issue of convincing (or admitting to) management that the business case for moving to the cloud was wrong. Second, like Royston, he noted there’s the question of how much talent and resources are needed to run on-prem infrastructure. Third, companies also have to consider the sensitivity of the data being moved and any security requirements that might be related to doing so. And lastly, Pichai said companies have to work through migration issues related to how best to transfer workloads that rely on specialized tools available in each cloud.  

“It’s simply not just about the data, it’s about all the add-on services that are provided in those kinds of environments,” he said. “You’ll find it incredibly hard to find equivalent services in the private cloud setup.” 

Linthicum concurred, stating that while some workloads may be repatriated, the public cloud will likely remain the right choice for most applications. 

Still, according to Pichai, hyperscalers are watching the repatriation movement closely. And already they’re setting themselves up to win no matter which route enterprise customers decide to take.  

“The hyperscalers are also being very smart about this because they’re bringing these kinds of services, saying ‘hey, if you want to run it on-premises, we can do that.’ Because it’s in their interest to have their software and their tooling in your IT environment,” Pichai said.  

Ultimately, Linthicum tipped the general trend of enterprise migration to the cloud to continue, even if some workloads and data stores end up being moved back on-premises.  

“Repatriation will be a thing, perhaps more popular than it was, as workloads are rightsized to more cost-effective platforms. However, the hyperscalers will continue to thrive because they want their customers to be successful with any platform that they choose, and that should be the end game here,” he concluded.