A quiet but consequential reversal is underway in enterprise technology strategy. After a decade in which ‘cloud-first’ was the default posture for digital transformation, organizations are now recalibrating. A recent Cloudera survey found that 66% of respondents moved AI workloads from public cloud back to private cloud or on-premises environments over the past 12 months. Broadcom’s Private Cloud Outlook 2026 corroborates this trend, reporting that 56% of enterprises are running or planning production AI inferencing on private cloud, with 83% actively considering or already executing repatriation. For CFOs, General Counsel, and CTOs, this is not a technical footnote — it is a strategic inflection point with direct consequences for cost structure, risk exposure, and M&A due diligence.

The Repatriation Data: Why Public Cloud No Longer Fits Every AI Workload

The economics and governance profile of AI have changed the cloud calculus. Training and running large language models generates unpredictable, often escalating compute costs that are difficult to forecast under public cloud consumption pricing. Simultaneously, AI workloads increasingly touch sensitive data — financial records, IP, customer information, HR data — that boards and regulators expect to remain under demonstrable organizational control. Repatriation is therefore less about rejecting cloud migration outright and more about matching infrastructure to workload sensitivity and cost predictability. For mid-market firms in particular, this represents an opportunity to scale AI adoption in enterprise operations without inheriting the compliance and cost exposure of hyperscaler-dependent architectures.

This is a maturing of digital strategy, not a retreat from it. Enterprises are moving from a one-size-fits-all cloud model toward a deliberately hybrid architecture — public cloud for elastic, non-sensitive workloads; private or on-prem infrastructure for regulated, high-value AI use cases.

Europe’s Sovereignty Imperative: Compliance, Power, and Infrastructure Planning

The European dimension of this shift is distinctive. Reuters reporting on Europe’s cloud strategy highlights a wave of AI data-center expansion pushing outward from congested hubs like London and Frankfurt, driven by power availability and land constraints — a physical reminder that digital transformation has real-world infrastructure limits. ISG research further shows German enterprises actively redesigning private and hybrid clouds specifically for AI workloads, reflecting a European preference for control, auditability, and operational resilience that aligns closely with GDPR, the EU AI Act, and emerging data sovereignty expectations.

SAP’s cloud backlog rising 26% to €22.9 billion is a telling signal: European companies are not abandoning cloud migration, but they are consolidating core finance, procurement, supply-chain, and HR systems onto platforms explicitly built to support compliant AI deployment. This is innovation management done deliberately — sequencing modernization of the data and process layer before layering AI on top, rather than bolting AI onto legacy systems and inheriting their governance gaps.

Modernization of Core Systems as the Real Foundation for AI Adoption

The Hitachi Digital Services–adesso partnership, focused on accelerating AI, cloud modernization, and digital engineering, illustrates a broader market pattern: enterprise AI adoption is increasingly bundled with core system modernization rather than pursued as a standalone initiative. This has direct implications for M&A and transformation advisory work. Target companies with fragmented, unmodernized core systems face materially higher integration risk and slower time-to-value on AI initiatives post-transaction. Due diligence teams should now treat cloud architecture maturity and data governance readiness as material factors in valuation and integration planning, alongside traditional financial and legal review.

Implications for Business Leaders

  • CFOs should reassess cloud spend forecasts and build cost-of-ownership models that compare public cloud consumption against private infrastructure capex for AI-intensive workloads.
  • General Counsel should map which AI use cases touch regulated data and ensure infrastructure choices are defensible under GDPR and the EU AI Act’s risk-tiering framework.
  • CTOs should treat hybrid architecture as a design principle, not a transitional state, and build workload classification frameworks that route data by sensitivity and regulatory exposure.
  • M&A Directors should incorporate cloud and AI infrastructure maturity into due diligence checklists, given the direct link between core system modernization and post-merger integration speed.

Key Takeaway

The shift from public cloud to private and hybrid infrastructure marks a new phase of enterprise digital strategy — one where AI adoption in enterprise settings is inseparable from governance, sovereignty, and cost discipline. Organizations that treat this as an architectural decision, rather than a purely technical one, will be better positioned to scale emerging technology responsibly while satisfying regulators, boards, and cost-conscious stakeholders alike.