To migrate to the cloud or remain on prem?
What IT pros should consider when making decisions about whether to move to the cloud or keep data and workloads on premises.

Illustration: Shannon May, Photo: Adobe Stock
What IT pros should consider when making decisions about whether to move to the cloud or keep data and workloads on premises.

Illustration: Shannon May, Photo: Adobe Stock
To migrate to the cloud, or to remain on prem: That is the question vexing many IT pros today.
Just a few years ago, the conventional wisdom was to send just about everything to the cloud. Now, in the AI era and amid heightened data sovereignty concerns, the tides are shifting. Companies are pursuing hybrid strategies—or even conducting reverse cloud migrations.
“There are certain things that shouldn’t be in the cloud,” Mattias Fridström, VP and chief evangelist of Swedish telecom Arelion, told Morning Brew.
The cloud probably makes sense for orgs simply looking to store large amounts of data, he said, but those with more sophisticated needs should evaluate their options.
“The assumption that everyone made, that everything should be in the cloud because it’s cheaper, I think that’s the biggest mistake that most companies have [made],” Fridström, who oversaw a two-year migration of Arelion’s data to the cloud, added, “including ourselves.”
Benefits of cloud migration, or the process of moving workloads and data off-site and onto cloud infrastructure, include cost efficiency, flexibility, and potential for faster development cycles. But the process also comes with risks around data security, compliance, and latency, according to a blog post on developer platform Cortex.
Randy Armknecht, managing director and global infrastructure, cloud, and security engineering practice lead at global consultancy Protiviti, told us that customers are re-evaluating their cloud migration strategies because of AI. He’s seen this trend play out in the manufacturing industry, where workflows and data that were essential for production were kept on prem to begin with.
“But they would often collect that data, ship all the data off to the cloud, and run data analysis and other things in the cloud against that data to see how things were going,” Armknecht said. “With the advent of AI, there’s been more business cases where they’re like, ‘Hey, we have data on the plant floor, and if we use AI, we could get real-time adjustments.’ Think, like, visual defect-type detection. That’s not data that you necessarily want to send to the cloud and process remotely and then send back.”
As such, he’s seeing customers stock up on hardware and move workloads back onto the plant floor.
Armknecht doesn’t advise his clients to migrate to the cloud if cost is their primary reason. It often doesn’t end up being the cheapest solution, but there are other legitimate reasons to choose that approach, he said.
“If you write infrastructure as code and deploy it into, say, the East region, and then you want to have failover in the West region, you can make some updates to your code, you can deploy it to the West, and ta-da, you now have your environment in two locations,” he said. “Versus if we think about on prem, it’s like scout sites, sign a lease, ship hardware. It’s much more complicated, much more involved. So I think the cloud provides a lot of great operational benefits. And at the end of the day, you’re probably gonna pay more for that convenience and those benefits.”
Another factor to consider? Data sovereignty. Fridström said there’s been more concern about this amid geopolitical turmoil.
“Suddenly our own company and everyone else is starting to think about, ‘Do I really want my data in the cloud? And what type of data should absolutely not be in the cloud?’” he said. “While you previously trusted the cloud providers 100%...suddenly everyone wants to know where their servers are.”
Meanwhile, AI has introduced flexibility that makes it easier to keep certain workloads on prem.
“Even more companies are starting to think about pulling even more systems back to their own data centers, both from a security point of view but also from a usability point of view,” Fridström said.
He used inventory systems and financial data as examples of things that should likely stay on prem: “We don’t want that data to be somewhere else in the cloud, and someone else seeing that data.”
Cost doesn’t seem to be the No. 1 driver of decision-making these days, unlike a few years ago, Fridström said—nor should it be. Instead, organizations seem intent on finding the ideal solution for their needs, and then evaluating on a cost basis.
“When the CEO asks the head of IT, ‘Do you trust your data and do you have control of your data?’ you really want to say yes to that question,” he said. “And then you don’t want to say, ‘We’ve optimized the cost perfectly and I really have no idea where my data is but hopefully safe.’ I think those days are over.”
A mistake he sees organizations making is assuming that everything should be in the cloud because it’s cheap.
Armknecht, meanwhile, sees orgs misunderstanding a system’s full data flow when they’re making decisions around cloud migrations.
“They may look at or even think of a system as pieces A, B, and C, and they’ll say, ‘Hey, this is all cloud-ready, we can move it to the cloud,’” he said. “And what they didn’t realize is that…another piece of the team has parts D, E, and F, and maybe one of them’s in the cloud and the other two are on-prem, and now you have data and transactions flowing back and forth through the environment.”
He advised IT pros to make sure they’re aware of the workloads and computing needs of their orgs, because that can be helpful in determining the role of cloud infrastructure. It’s also important to assess whether your team is ready to secure a cloud environment, he said.
Chris Thomas, principal, Deloitte Consulting, and US hybrid cloud infrastructure leader, said via email that he would start the decision-making process with one question: “Am I putting the right workload in the right place, on the right model, at the right tier, and with the right resiliency?”
He pointed to “highly regulated and sovereign workloads, especially those with tight latency needs or heavy data gravity” as good candidates for staying in an on-site data center or on the edge.
And don’t go it alone, Fridström advised: “Talk to your network provider and talk to your storage experts.”