On-Prem Infrastructure Is Staging a Real Comeback, Driven by AI Costs and Cloud Security Fears

Quick Facts

• 86% of CIOs plan to repatriate some public cloud workloads in 2025, up 2x from the prior year, the highest share ever recorded.
• Companies running open-source AI models on-premises are cutting operational costs by up to 70% compared to cloud API-based solutions.
• 92% of IT leaders express confidence in on-premises cybersecurity, versus 78% for fully cloud-based environments.

The cloud is still growing. Global cloud spending is projected to hit $723.4 billion in 2025, up from $595.7 billion in 2024. But a parallel trend is accelerating: enterprises are moving workloads back on-premises at a record pace, and the reasons go beyond simple cost cutting.

According to data cited in a Crunchbase analysis by strategic adviser Itay Sagie, 86% of CIOs plan to move some public cloud workloads back to private cloud or on-premises environments in 2025. That figure is double the rate from the year before and the highest ever recorded. A separate survey found 87% of organizations plan to repatriate some or all workloads over the next two years.

Security Is the Top Driver

IT leaders trust on-premises systems more when it comes to security. In current surveys, 92% of IT leaders express confidence in on-premises cybersecurity, compared to 78% for fully cloud-based environments.

The July 2024 CrowdStrike outage sharpened that concern. A faulty software update crashed approximately 8.5 million Windows systems globally, grounding flights, halting hospital operations, and taking banks offline. The incident exposed how tightly coupled cloud-dependent architectures can fail at scale.

CrowdStrike CEO George Kurtz has since warned that the threat environment is intensifying. “AI is rewriting the enterprise attack surface at breakneck speed,” Kurtz said. “Each prompt becomes an entry point for the adversary.”

AI Is Accelerating the Shift

The rise of capable open-source AI models is giving companies a new reason to bring workloads in-house. Idit Levine, founder and CEO of Solo.io, says her enterprise customers are asking a practical question after experimenting with major model providers: “Can I take an open source model and run it on-prem? It will do almost 90% of what the big one’s doing. It will cost way less.”

The numbers support that logic. Organizations running open-source AI models on-premises report cutting operational costs by up to 70% compared to API-based cloud solutions. Cloud waste hit 29% in 2026, and nearly 95% of IT leaders have encountered unexpected cloud charges that disrupted budgets or slowed projects, according to a 2025 Backblaze survey.

The Cost Case Is Real

Several high-profile examples illustrate the financial argument. One company reduced annual infrastructure costs from $3.2 million to $720,000 after establishing on-premises infrastructure with a $600,000 upfront investment. 37signals reported saving $1 million per year by moving email and other services back on-premises. Dropbox saved hundreds of millions by building its own storage infrastructure after outgrowing public cloud providers.

Basecamp projected $7 million in savings over five years by avoiding cloud lock-in. The UK Cabinet Office estimated that overreliance on a single cloud provider could cost public bodies 894 million pounds.

Regulation Is Forcing the Issue

Compliance requirements are removing optionality for companies in regulated industries. The EU’s Digital Operational Resilience Act took effect in January 2025, requiring financial firms to demonstrate full control over their technology architectures, including subcontracting arrangements. The EU AI Act reaches full application in August 2026.

GDPR, the EU Data Act, NIS2, and sector-specific rules demand tighter control over data flows and limit dependence on foreign cloud providers. For many organizations in finance, healthcare, and government, keeping data on-premises is no longer a preference. It is a legal requirement.

What This Means for SaaS Vendors

The shift creates direct pressure on SaaS companies built entirely on public cloud delivery. Customers in regulated industries or those managing sensitive AI workloads are asking harder questions about data residency, portability, and vendor dependency. SaaS providers that cannot offer on-premises or private cloud deployment options risk losing enterprise deals to vendors that can.

42% of companies are already considering moving workloads back on-premises specifically to escape vendor dependencies. That number will climb as open-source AI models continue to improve and regulatory pressure grows across major markets.

Read more: Is On-Prem Making A Comeback?