SAP Emphasizes European AI Sovereignty at Sapphire

Per Forbes and ERP Today, SAP's Sapphire Madrid keynote repeated its Sapphire Orlando content with one notable change: where Anthropic had a speaking slot in Orlando, Mistral and n8n were given stage time in Madrid to present their capabilities. Philipp Herzig, SAP's chief technology officer, told the Madrid audience the company is "taking a look at sovereignty end-to-end with different levels of control and resilience" before introducing SAP's European AI partners. In a press Q&A, CEO Christian Klein said many of SAP's European customers - particularly in the public sector and regulated industries - are "concerned about the geopolitical situation in the world" and worry about scenarios such as sanctions blocking data transfers between countries or a jurisdiction barring access to the latest version of a given large language model: "You never know what might happen tomorrow." Herzig added that ongoing deglobalization is driving the demand, and that sovereignty concerns extend beyond Europe to Canada, Singapore, Australia and other countries. Klein was direct about the limits of the pitch: "You will never have complete independence on the hardware side. But where it really matters is that the data is stored here and that there are European AI companies."

SAP's EU AI Cloud, introduced in 2025, gives European enterprises and German public-sector customers GDPR- and EU AI Act-compliant access to a choice of LLMs and data centers, with data that stays in Europe and is never used to train third-party models, according to SAP's own product page. Mistral AI and Cohere Enterprise are its featured model partners; the full Mistral platform is now generally available on SAP's Business AI Platform with sovereign model access, and n8n provides the visual AI workflow orchestration layer inside Joule Studio, SAP's environment for building and governing custom agents. ERP Today reports SAP frames sovereign deployment as a tier structure: secure public cloud, SAP-operated sovereign cloud run within a customer's region under local rules, and more tightly controlled environments for government and classified workloads.

SAP COO Sebastian Steinhaeuser described a staged-autonomy control model for agents operating in finance, procurement, supply chain, HR and other processes: customers start with a human in the loop, verify agent outputs, and expand autonomy as confidence builds, per ERP Today. Herzig separately distinguished probabilistic AI use cases from mission-critical, low-error-tolerance processes such as financial postings or payroll, which SAP is engineering with deterministic guardrails rather than treating as generic model calls. SAP is also testing physical AI in production: ERP Today reports SAP has deployed Cyberwave robots in its own logistics warehouse in St. Leon-Rot, Germany, handling live box-folding, packaging and shipping, connected to SAP Logistics Management, SAP Business Technology Platform and SAP's Embodied AI Service - an early signal that agentic AI at SAP is starting to extend from software workflows into physical operations.

Whether other enterprise software vendors respond with their own named regional model partnerships and tiered sovereign-cloud offerings; actual customer adoption of EU AI Cloud and any published data-residency certifications; whether the Cyberwave warehouse deployment expands beyond a single pilot site; and how sovereignty demand plays out in the non-European markets Herzig named, including Canada, Singapore and Australia.

Vendors facing deglobalization pressure in regulated markets typically respond by combining three things: named local model partners, data-residency guarantees, and governance tooling for auditability - a pattern SAP is following closely here. SAP's own AI innovation lead, Jonathan von Rueden, told Forbes that European partners Mistral and n8n are not as capable as Anthropic by some measures but are meaningfully cheaper, and that model leadership shifts quickly enough that which vendor currently leads may matter less than the governance and deployment architecture wrapped around it - a framing that is itself a sales pitch for SAP's platform-level lock-in as much as it is commentary on model quality.