AI Is Forcing Businesses To Rethink Their Data Strategies

How AI Is Changing Data Strategies

AI is the catalyst for this rebalancing. Many organizations begin AI initiatives in the cloud because it offers fast access to computing resources and avoids upfront capital expenditures. Proofs of concept are easier to launch in elastic cloud environments, particularly for organizations that haven’t yet modernized their on-prem infrastructure.

But once those AI projects prove valuable and move into production, the calculus often changes.

“AI outcomes have an intense resource requirement,” Brodsky says. “When you look at cloud consumption models, it isn’t always the most financially effective way to run those workloads long-term.”

Performance is another driver. AI workloads that require low latency — such as real-time analytics, inference or data-heavy processing — often perform better when data and computing are located close to the business. “If timing is critical, you want your data as close to your business as possible,” Brodsky says. “That’s where cloud isn’t always the ideal fit.”

Security and governance concerns also loom large, particularly in regulated industries such as healthcare, financial services and the public sector.

UP NEXT: Businesses must personalize their AI to remain competitive.

On-Prem Hardware’s Advantages

For many organizations, repatriation is closely tied to data governance. Keeping sensitive data on-premises can provide clearer visibility into where data lives, who can access it and how it’s protected. That’s important from a data security standpoint in general, but those considerations are especially vital for meeting regulatory and compliance requirements.

“When it’s on-prem, you have control,” Brodsky says. “But you also have the risk.”

That trade-off is critical. Rashid Rodriguez, a cyber resiliency practice lead at CDW, notes that public cloud platforms include built-in protections — such as geographic redundancy and disaster recovery — that don’t automatically carry over when workloads move back on-premises.

“In the cloud, disaster recovery is largely baked in,” Rodriguez says. “When companies bring workloads back, they need to make sure they’re not losing the level of protection they previously relied on.”

That often means rethinking backup strategies, recovery architectures and cyber resilience more broadly. Rodriguez points to growing interest in more advanced approaches, including immutable backups and orchestrated recovery models that can support mission-critical AI workloads.

Repatriation Is Not a Lift-and-Shift Proposal

One of the biggest misconceptions about cloud repatriation is that it’s a simple reversal of a cloud migration. In reality, AI workloads frequently exceed the capabilities of existing on-prem infrastructure.

“Servers that were procured three years ago may not be able to handle what these applications require,” Brodsky says.

As a result, repatriation decisions often trigger broader modernization efforts, including new hardware, increased power and cooling capacity, and redesigned architectures. Before making those investments, organizations need a clear understanding of their current environment and future requirements. Rodriguez says that typically starts with assessments that examine infrastructure scalability, security tooling, licensing, backup capacity and disaster recovery readiness.

“You have to evaluate whether your on-prem environment can actually ingest and protect what you’re bringing down from the cloud,” he says.

FIND OUT: How next-gen AI is impacting cybersecurity.

Timelines and approaches vary. Some organizations opt for high-level assessments to guide strategy, while others pursue deeper technical workshops or phased transitions based on business priorities and service-level agreements.

Despite the renewed interest in on-prem infrastructure, cloud repatriation doesn’t signal a retreat from cloud computing. Instead, it reflects a more mature understanding of hybrid IT.

“Five years ago, we had daily conversations with customers who wanted to be 100% cloud,” Brodsky says. “Very few actually got there.” Today, most organizations operate hybrid environments by necessity, balancing cloud flexibility with on-prem performance, cost predictability and governance. AI is simply forcing a renewal of those considerations.

As businesses navigate those decisions, many turn to partners with experience across both cloud and on-prem environments to help assess options, plan transitions and modernize infrastructure where needed. “These conversations aren’t yes-or-no decisions,” Brodsky says. “They’re consultative by nature, because everything is interconnected.”