Sovereign connectivity for medical research: Apheris and T-Systems enable the secure use of clinical data

10-20-2026 4 min read

• Patient data remains at the hospital; AI computing capacity from Germany
• Apheris networks for pharmaceutical and clinical research run on the Industrial AI Cloud
• Already tested: a clinical network with data from more than 50,000 patients

Hospitals have valuable treatment data. However, using it for studies across multiple sites often requires substantial effort. Together, Apheris, an AI company serving the biotechnology and pharmaceutical industries, and T-Systems aim to make this easier. Apheris operates federated data networks for drug discovery and clinical research. These networks enable participants to train models and conduct analyses across protected datasets. The underlying data remain protected and within each partner's own environment. NVIDIA supports this work through collaboration on AI models, software, and accelerated computing.

Apheris and T-Systems now enable their customers to operate Apheris’ federated networks for drug discovery and studies using healthcare data in Europe on the sovereign infrastructure of T Cloud Public. T-Systems’ Industrial AI Cloud (IAIC) provides European customers with the GPU resources required for this. The partners will present details of their collaboration at the NVIDIA GPU Technology Conference (GTC), taking place in Berlin from October 20 to 22, 2026.

Why studies involving clinical data are complex

The General Data Protection Regulation and other data protection requirements call for clear rules governing access to and processing of treatment data in multi-site studies. Internal policies and contracts must also be aligned. As a result, approvals and technical preparations can make new research projects complex and time-consuming.

Double protection for sensitive data

Apheris relies on federated learning: The model is brought to the data and trained locally in each hospital's environment. Patient data stays on site. The privacy-compliant model parameters trained locally are transmitted to T-Systems' Industrial AI Cloud. There, Apheris combines them into a global model. The cloud, which is operated in Germany, uses NVIDIA GPUs.

Using health data securely

"Health data is among the most sensitive data of all. Anyone conducting research with it needs reliable, secure, sovereign infrastructure," says Ferri Abolhassan, CEO of T-Systems. "With Industrial AI Cloud, we offer the right solution. Together with Apheris, we enable medical progress without sensitive data leaving the hospital."

"Patient data must remain protected, which is why research can barely use it today. In a world where AI drives progress, it is crucial to make this data usable without giving up patient protection," says Robin Röhm, CEO and co-founder of Apheris. "With federated learning, we bring the model to the data so it can learn from it while the data remains protected where it is. With T-Systems, this now also takes place on sovereign European infrastructure." Proven in practice

According to Apheris, a healthcare data network connects several hospitals and includes data from more than 50,000 patients, including electronic health records and imaging data. Research teams can conduct analyses directly at the respective sites without centrally pooling patient data.

In a feasibility study published in 2025, researchers examined the segmentation of MS lesions using a total of 512 MRI cases from three clinical institutions. The imaging data remained at the respective sites.

Also for drug discovery

In pharmaceutical research, companies use Apheris to train AI models together without disclosing their data. In the AI Structural Biology (AISB) Network, nine leading pharmaceutical companies work together on AI for structural biology. Other Apheris networks improve predictions of small-molecule properties (ADMET) and antibody developability.

Apheris uses NVIDIA FLARE as its core framework for federated learning and contributes fixes for production deployment to the project. The company uses NVIDIA cuEquivariance kernels to accelerate training and inference for supported workloads, including OpenFold3. In addition, the ApherisFold platform uses Nvidia MSA NIM, enabling accelerated MSA generation.

About Apheris

Apheris operates large federated data networks for the life sciences industry and uses them to train AI models for drug discovery and clinical research. The networks cover co-folding, binding affinity, ADMET, in vivo PK, and antibody developability, as well as Real-World Evidence (RWE) based on patient data. Within these networks, leading pharmaceutical companies and healthcare organizations train models on proprietary, distributed data that never leaves its environment, helping protect data, intellectual property, and patient privacy. The result is models that can outperform publicly available alternatives, especially where those alternatives are limited by insufficient data. They are delivered through Foundry, an AI-first platform embedded directly in partners' research workflows as modular building blocks that agents can call.