SAP Exit Strategies for the Cloud and AI

SAP ERP Needs a Plan B for Business AI

This unprecedented intervention exposes the vulnerability of European companies that have blindly relied on American IT giants as part of their digital transformation. For the Walldorf-based software company SAP, this export control strikes right at its strategic Achilles’ heel, as SAP CEO Christian Klein had only recently selected Anthropic’s language model Claude as the primary “reasoning model” for its digital assistant Joule and the entire Business AI Platform.

The much-touted vision of the Autonomous Enterprise—in which AI agents are expected to autonomously handle business-critical tasks such as quarterly financial closings or supply chain management in the future—is thus rooted in a geopolitical fault line. The risks for existing SAP customers are glaring: If a foreign government can shut down the cognitive engine of a European ERP system overnight, there is a risk of an abrupt loss of functionality and an incalculable operational downtime risk for the entire value chain.

SAP vendor lock-in

This dangerous trend is forcing ERP decision-makers to radically rethink and expand their defensive strategies. While the user association DSAG and critical existing SAP customers have for years been vehemently calling for a legally and technically sound cloud exit strategy to escape vendor lock-in from contracts such as Rise with SAP, the urgent need for a dedicated SAP AI exit strategy is now becoming apparent.

The deep integration of external large language models (LLMs) carries the risk of massive business, organizational, and technical dependence, as applications are often tailored to the specific API structure, prompt format, and output behavior of a particular AI provider. If this specific model were to suddenly become unavailable due to political sanctions or strategic restrictions, the painstakingly developed AI processes could become worthless; therefore, having a prepared, architectural exit and migration strategy for AI services will become a matter of survival for every company in the future.

Exit: SAP Business Technology Platform with GenaAI Hub

Paradoxically, it is the Business Technology Platform (SAP BTP)—which many customers often view with skepticism—that offers a lifesaving architectural escape route during this crisis to compensate for the loss of Anthropic. With the BTP’s Generative AI Hub, SAP has established an abstracting intermediary layer (adapter pattern) that sits between the ERP application and external AI providers.

In theory, this BTP architecture allows users to quickly replace blocked models and flexibly redirect their prompts to other providers such as Microsoft Azure/OpenAI, Google Vertex AI, or to open models such as Meta’s Llama, Mistral AI, or IBM Granite. However, to truly take advantage of this opportunity, developers at the SAP base level must maintain the highest level of architectural discipline and strictly avoid any hard-coded dependencies on the proprietary formats of a single LLM.

The EU Data Act as an Exit Strategy

At the regulatory level, the Anthropic lock-in also highlights the massive gaps in current European legislation. The EU Data Act is currently attempting to break cloud lock-in and provide legal safeguards for cloud exits by requiring providers to ensure data portability and remove technical barriers. However, this framework falls far short in the age of agentic AI, because the purely physical handover of raw data is effectively worthless to a company if the associated cognitive application logic is removed.

The European Union will therefore have no choice but to massively expand its regulatory framework beyond the EU AI Act and the Data Act in order to enshrine binding AI exit strategies and strict interoperability standards for foundation models in law, so that European economic systems no longer remain hostage to foreign monopolies.

The mood within the SAP community in the wake of this AI disaster is a volatile mix of validation and deep concern. Long before the U.S. ban, there was widespread skepticism among SAP users regarding the AI promises made by the Walldorf-based company. The latest DSAG Investment Report confirms that 77 percent of SAP customers who are already using AI scenarios in production are deliberately doing so with non-SAP solutions and avoiding the expensive, restrictive SAP AI ecosystem. The blocking of Anthropic confirms the fears of these critical users and fuels the vehement demand for true digital sovereignty. The call within the community for hybrid architectures—in which core SAP data never leaves the company and local open-source LLMs are instead run on the company’s own enterprise hardware—is growing ever louder. For existing SAP customers, the Anthropic block is the final proof that the blind path toward U.S.-dominated cloud and AI dependency is a strategic dead end from which they can only free themselves through consistent, technical self-reliance.