Executive Summary
Cloud migration in German enterprises has evolved from a strategic project to fragmented infrastructure. While 90 percent of German companies use cloud services and 41 percent employ multi-cloud, disillusionment is simultaneously growing: cost control, skill shortages, and sovereignty concerns stem from often unplanned complexity. The integration of AI workloads significantly exacerbates these challenges and requires a shift in thinking from administrative to engineering-driven governance. Multi-cloud is no longer merely an infrastructure decision, but a strategic business risk management task with geopolitical dimensions.
People
- Tobias Regenfuss (Accenture, Senior Managing Director Technology)
Topics
- Multi-cloud architecture and governance
- Cloud costs and FinOps
- Digital sovereignty and data protection
- AI infrastructure and GPU computing
- Skills management and platform engineering
Clarus Lead
German enterprises are experiencing the downside of cloud euphoria: strategic migration projects have become an uncontrolled multi-cloud patchwork that endangers costs, security, and sovereignty. 86 percent of CIOs worldwide are already planning to repatriate workloads from Public Cloud – a record high. The central insight: multi-cloud requires not better management, but fundamental engineering thinking with clear strategies for data flow, specialization, and geopolitical risks.
Detailed Summary
The German corporate landscape reveals a paradox: while cloud adoption is officially successful, significant problems emerge in operational reality. 41 percent of companies rely on multi-cloud, yet 82 percent simultaneously demand German or European hyperscalers as alternatives to US providers. This is not contradictory – rather, it reveals deep unease with the current architecture. Many infrastructures labeled "private cloud" are actually renamed legacy data centers without genuine cloud-native properties.
Complexity emerges in three underestimated dimensions: First, the technical landscape itself – from Kubernetes through service meshes to specialized AI hardware, the number of required competencies explodes exponentially, while organizations typically learn only linearly. Second, the cost trap: the original promise of "pay-as-you-grow" transforms into "pay-and-lose-track." Unused reservations, opaque egress costs, and GPU resources for AI training cost multiples of classical compute capacity. 89 percent of companies report intransparent cost structures. Third, geopolitical fragmentation: 78 percent of German companies see critical dependency on US cloud providers. Regulatory requirements such as NIS2 and DORA intensify this pressure further.
The solution lies in five critical success factors: strategic clarity with rigorously defined scope, concrete sovereignty based on realistic risk analysis per workload, engineering excellence through Platform Engineering, central control plane as governance centerpiece, and pragmatic evaluation of European options such as STACKIT or T-Systems Sovereign Cloud – not from idealism, but based on genuine requirements.
Key Takeaways
Strategic failure precedes technical: Unplanned multi-cloud environments arise from missing business clarification, not from lack of technology.
AI amplifies everything: Specialized hardware, model lifecycle management, and GPU costs make multi-cloud management exponentially more complex.
Sovereignty is differentiated: Not all workloads require European infrastructure – honest risk analysis per use case is required, not blanket demands.
Control plane is decisive: Central governance layer for service catalog, monitoring, and FinOps is prerequisite to keep complexity manageable.
Skills are the bottleneck: Investments in training, recruiting, and external expertise are not optional, but existential for multi-cloud success.
Critical Questions
Evidence: What specific measurement criteria underpin the Bitkom statistics (41% multi-cloud usage)? Is this actually measuring multi-cloud or merely multiple cloud services without genuine architectural integration?
Data Quality: The article references the Barclays CIO Survey 2024 (86% workload repatriation). How representative is this global sample for German companies with different regulatory requirements?
Conflicts of Interest: The author is a senior technology executive at Accenture, a consulting firm that benefits from more complex multi-cloud requirements and associated consulting services. Might complexity be overstated to justify consulting demand?
Causality: The text directly connects AI workloads to multi-cloud problems. Are AI requirements truly a primary driver of cost uncontrollability, or rather a symptom of missing governance that would exist with traditional workloads as well?
Implementation Risks: The call for "engineering excellence" and specialized expertise assumes resources the text identifies as tightly constrained (skills gap). How realistic is this recommendation for mid-sized companies without global IT budgets?
Lock-in Paradox: The text warns against commercial lock-in with hyperscalers, yet simultaneously recommends central control planes (e.g., Red Hat OpenShift). What lock-in risks emerge from proprietary governance platforms themselves?
European Alternatives: Only 12% of German companies would use European cloud services if functional disadvantages resulted. Does this not refute the fundamental premise of sovereignty necessity?
Measuring Sovereignty: The text demands concrete risk matrices for sovereignty, but offers no framework for such measurement. How does a CIO operationalize "protection against data exfiltration" or "safeguarding against geopolitical lock-outs" concretely?
Source List
Primary Source: Why Multi Cloud is the Operating System for the AI Era – computerweekly.de
Cited Studies and Data Sources (from article):
- Bitkom Cloud Study 2025 (90% cloud usage, 41% multi-cloud, 78% sovereignty concerns, 12% acceptance of European delays)
- Barclays CIO Survey 2024 (86% workload repatriation from public cloud)
- Global enterprise studies (89% intransparent cost structures)
Author: Tobias Regenfuss, Accenture (Senior Managing Director Technology for Central and Eastern Europe)
Verification Status: ✓ February 19, 2026
This text was created with the support of an AI model.
Editorial responsibility: clarus.news | Fact-check: February 19, 2026