Healthcare AI often stalls between a promising pilot and the data controls required for clinical operations. SAP and Fresenius plan to address that gap with a sovereign platform designed to connect hospital systems and run models inside a controlled European environment.
The companies intend to invest a mid-three-digit million-euro amount over the medium term. SAP contributes Business AI and Business Data Cloud, while Fresenius brings healthcare operations and clinical expertise.
The ambition is not one medical model. It is a reusable digital backbone that makes sensitive data available to approved applications without surrendering sovereignty, interoperability, or accountability.
Interoperability Is the First Infrastructure Problem
Hospitals operate varied information systems, electronic medical records, and specialist applications. Fragmentation forces each AI project to create new connections and makes it difficult to understand where patient information moved.
SAP’s AnyEMR strategy uses open standards including HL7 FHIR to connect hospital information systems, EMRs, and medical applications. A common layer can provide consistent access while allowing institutions to retain systems that cannot be replaced quickly.
Interoperability must include meaning as well as transport. The same clinical concept can be recorded differently across organisations, and missing context may turn a technically valid data exchange into an unsafe model input. Shared definitions and quality controls belong in the backbone.
FHIR can provide a common exchange format, but implementation choices still differ. The partners will need conformance testing, terminology mapping, and visible data-quality warnings so an application knows when two apparently similar fields should not be treated as equivalent.
Sovereignty Creates a Controlled Place for Models
Public cloud services may not provide the jurisdictional, contractual, or operational control European healthcare organisations require. The proposed platform is intended to keep processing compliant and govern which models may use which information.
Sovereignty is more than locating servers in Europe. Hospitals need control over encryption keys, administrators, subcontractors, model updates, retention, and secondary use. They must be able to trace access and remove a service without losing their clinical data.
This foundation could let Fresenius build applications spanning the care chain while giving doctors and hospital teams simpler, scalable assistance. The platform should make human responsibility visible; a sovereign model can still be wrong.
Hospitals must also preserve operational choice. Contracts should define how data and applications move if a provider changes, how outages are handled, and which controls remain under the institution’s direct authority. Sovereignty loses meaning if technical lock-in prevents practical control.
An Ecosystem Expands Value and Supply-Chain Risk
SAP and Fresenius plan joint investment in startups and scaleups alongside internal development. A broader tool library can accelerate specialised use cases, because no single company will build every clinical workflow.
Each additional provider introduces dependencies. Entry requirements should cover security, evidence, interoperability, data use, financial stability, and incident response. Applications need lifecycle plans so a hospital is not trapped when a startup changes direction.
The open ecosystem also needs consistent evaluation. Administrative automation, clinical research, and patient-facing recommendations carry different risk. The platform should apply controls proportionate to intended use rather than treating every AI service alike.
Michael Sen of Fresenius frames the goal as making data and AI everyday companions that create more time for patient care. Christian Klein of SAP emphasises sovereignty, security, and innovation. Those aims reinforce each other only if the infrastructure reduces integration work without weakening professional oversight.
The partnership suggests Europe’s healthcare AI race may be decided at the platform layer. Models will continue to improve, but hospitals need a controlled place to connect, validate, and operate them. A sovereign backbone can unlock that work if open standards and clinical accountability remain as important as national control.
