The great repatriation? IT leaders reset cloud strategies to optimize value

IT leader and former CIO Stanley Mwangi Chege has heard executives complain for years about cloud deployments, citing rapidly escalating costs and data privacy challenges as top reasons for their frustrations.

And he has seen some respond by moving workloads from the cloud back to on-premises data centers, a move they feel gives them better control over costs and privacy standards.

He worked with one insurance company that in 2023 made such a move, driven specifically by the desire to have a firm hold on its regulated data, for example.

Chege, now CEO and principal consultant of Digital Transformation Experts, says he has worked with other companies that have made similar moves. They, too, were motivated by data privacy issues, cost considerations, compliance concerns, and latency issues.

He cites the repatriation work he did with one bank several years ago, a project that shows how such decisions are often strategic and well thought out.

In that case, the board’s bank was troubled by the volume of unexpected fees charged by cloud providers. It also felt the provider couldn’t adequately help the bank meet the requirements of the Kenya Data Protection Act and the Central Bank of Kenya’s own regulations. And it believed the latency experienced during transactions hurt customer service.

“The bank decided it was better to be on-premise for certain workloads, where the cost-benefit analysis and total cost of ownership was going to be better in the long run,” says Chege, a member of the Emerging Trends Working Group with IT governance association ISACA.

Cloud workloads on the move

Such repatriations are not the norm, but they are not rare, either. Research shows that CIOs have been moving workloads back from the cloud for many years and continue to do so.

While some repatriations — particularly in the early days of cloud computing — resulted from failed cloud deployments, analysts and IT leaders say many repatriations today are smart moves that deliver benefits to the organization.

Moreover, these repatriations show how CIOs have a shrewder, more fluid cloud strategy today to ensure they don’t settle for less than what they want.

“Repatriation is a good option to keep,” says Natalya Yezhkova, research vice president at research firm IDC. “CIOs should be reassessing whether the public cloud is delivering [value], because the needs of workloads change, regulations around workloads change, offerings change whether in price or in functionality. So organizations shouldn’t close the door to either option, public cloud or a dedicated environment.”

Maintaining that credible threat of defection remains an important strategy for avoiding cloud vendor lock-in, especially in an era when the rapid emergence of AI tools brings a new twist to the opportunity costs of being stuck in a specific cloud.

Cloud repatriation: A consistent practice borne of common concerns

According to IDC’s June 2024 report “Assessing the Scale of Workload Repatriation,” about 80% of respondents “expected to see some level of repatriation of compute and storage resources in the next 12 months.” That 80% is consistent with past survey findings.

The report, which Yezhkova authored, notes that workload repatriation planning slowed in 2023 compared to 2022 but that the six-month period between the two most recent surveys (September 2023 and March 2024) saw increased levels of repatriation plans “across both compute and storage resources for AI lifecycle, business apps (CRM, ERM, and SCM), infrastructure, and database workloads.”

It noted decreased repatriation plans for engineering/technical workloads.

“We started to see repatriation from the moment public cloud was mainstream,” Yezhkova says. “There’s always some activity around repatriation, and the level of activity isn’t dropping off.”

IDC research also offers insights into why repatriation happens.

Poor planning frequently necessitated repatriation of workloads in the early years of public cloud adoption, Yezhkova says. As a result, organizations were unprepared to successfully optimize or even adequately run their cloud deployments and manage costs, prompting their move back to on-prem.

Now most organizations are capable of successfully managing their cloud deployments, she says, but they’re identifying security, privacy, performance, management, and governance needs that they have determined are better handled in a dedicated environment.

As an example, Yezhkova points to the complex considerations that organizations face when using generative AI, explaining that they are weighing the elasticity of the public cloud against the expanding costs that come with AI’s inexhaustible computing needs as well as the risks with exposing protected data to public AI models.

Multiple options in play for retreating from the public cloud

Cloud repatriation is typically defined as moving applications, data, workloads, or other resources hosted in the public cloud back to an on-prem data center or other dedicated environment (i.e., a private cloud).

IT execs now have more options beyond their own data centers and private clouds, namely as-a-service (aaS). For example, Hewlett Packard Enterprise’s GreenLake provides cloud-like flexibility to on-prem data centers and other dedicated IT environments, and Dell APEX Private Cloud offers an on-premises cloud experience for VMware workloads in data centers and edge locations.

Such products offer infrastructure that operates in ways similar to the public cloud and include a marketplace of applications and metering systems, Yezhkova says, adding that “these models make it easier to repatriate.”

CIOs also now can consider edge computing and micro data centers as alternatives to traditional dedicated data centers, cloud, and aaS models.

In addition to repatriation, CIOs can consider as alternatives to all-in public cloud the use of content delivery network services (such as those offered by specialty providers like Cloudflare) and GPU cloud services, says Brian Alletto, a director with digital services firm West Monroe.

“There are now options that give cloud-like experiences even if they’re not public cloud,” he says.

IT leaders and advisers say the availability of such options influences CIO decisions to move workloads out of the public cloud.

The IDC report addresses interest in these options, saying, “Demand for as-a-service (aaS) consumption of infrastructure resources across shared/public cloud and dedicated environments that may span on-premises datacenters, colocation facilities, and hosting sites is growing and becoming an essential part of hybrid multicloud strategies. Service-based consumption of compute/storage resources on-premises is still a new concept for enterprises, but awareness is growing. Availability of knowledge base and best practices on utilization of various aaS models will expedite the adoption process.”

Yezhkova says research shows CIOs are not repatriating entire systems or applications but rather specific workloads. According to the IDC repatriation report, less than 10% of respondents said they had repatriated entire workloads.

She further notes that the prevailing repatriation project involves a workload from a system that is deployed and owned by the organization but managed by a third party.

Cloud still reigns

Although research shows that repatriation continues, Alletto, Chege, Yezhkova, and others say it in no way suggests a pullback from the public cloud overall.

They each point out the challenges CIOs often encounter when repatriating workloads, saying that they could face the same problems they’re trying to eliminate with the move out of the cloud — specifically higher costs, security concerns, privacy problems, and performance issues.

“There can be a number of challenges with reverse cloud migration,” says Delie Minaie, senior vice president for digital and cloud solutions at Booz Allen Hamilton, citing data migration complexity, downtime and service disruption, security and compliance requirements, and application refactoring as typical pain points when repatriating.

“I’d caution [enterprise] leaders to think carefully before considering cloud repatriation,” Minaie says.

Chris Thomas, a principal at Deloitte Consulting and Deloitte’s US cloud leader, says enterprise leaders need to ensure they don’t fall victim to seeing an either-or choice when evaluating options.

He says companies are contending with a dynamic cloud landscape and they often look to minimize costs as their infrastructure grows.

“When looking to streamline costs, it is a natural reaction to say, ‘What can we take off of the cloud?’ But this implies a binary way of thinking: We can either migrate to the cloud or migrate off it,” Thomas says. “That isn’t always the case. There are optimization opportunities for companies who have already lifted and shifted to the cloud.”

He says those “optimization opportunities” include removing duplicate data sets, turning down unused instances, increasing automation and self-healing, investing in FinOps tools, and adopting new services offered by cloud providers.

Hybrid is the goal

Such calculations are happening, other experts say.

Indeed, Alletto says he sees more CIOs thoroughly evaluating their options. So when they do repatriate workloads, they are increasingly doing so because they expect benefits — and not because they were unsuccessful with a cloud deployment.

“The repatriations for good reasons are a higher percentage of repatriations now,” Alletto adds.

He points to his work on a recent merger, where the lead company had a goal to trim costs. That company determined it could move a workload out of the cloud and into an on-prem data center owned by the acquired company to get the desired savings without impacting performance.

“It shows that there are good reasons for repatriation,” Alletto says.

IT leaders say they are adjusting their cloud strategies to incorporate that perspective and allow for more options. They say there is a philosophical movement away from 100% cloud to hybrid as the ideal, with strategies that emphasize assessments of where workloads should be based on changing considerations.

“We keep tweaking to get the right balance, the right equilibrium,” Chege says.

Tej Patel, vice president for IT and CIO at Stevens Institute of Technology, is taking such an approach. He says: “The question should be: Where do I want to put my workload?”

Like others, Patel recognizes the benefits of public cloud, such as its scalability and elasticity to its numerous features and capabilities. But similar to others, he also acknowledges the challenges of managing and optimizing its use — specifically controlling costs and ensuring compliance with security and privacy standards.

As such, he has a hybrid strategy, seeing hybrid as the destination and not merely a way station to all-in with cloud. In fact, the university recently deployed high-performance computing (HPC) clusters on premises, after determining on-prem to be better than a public cloud deployment.

“We were very thoughtful in our approach,” he says. “We consider each scenario separately to make sure the infrastructure meets the needs of the user or the business services they’re looking for. The infrastructure exists to meet what the business wants to do. A hybrid environment can help deliver this.”