German Companies Shift AI Workloads from Public Cloud to On-Premises

A recent Enterprise AI Infrastructure Survey found that 79 % of German enterprises have already moved AI workloads from public cloud services back to their own on‑premises infrastructure, and the figure rises to 93 % when including those planning or currently executing the transition. The shift is driven by concerns over data sovereignty, unpredictable cloud costs and the need for real‑time performance, especially for retrieval‑augmented generation (RAG) applications.

Regulatory pressures amplify the trend. EU directives such as the NIS2 directive, GDPR, the EU AI Act and sector‑specific rules in health and finance require clear evidence of data location and limit certain cross‑border cloud arrangements. In addition, the US CLOUD Act presents a legal risk: U.S. authorities could potentially access data stored on American cloud providers regardless of where the servers are physically located, prompting German firms to reassess compliance and jurisdictional exposure.

Together, the findings underscore a broader re‑evaluation of cloud strategies among German businesses, emphasizing greater control, legal certainty and security over pure cost or scalability benefits.