Healthcare federated learning infrastructure
Microsoft Azure and NVFlare are nearly even for healthcare federated learning infrastructure.
It depends
Owkin’s answer focuses on pharmaceutical and academic medical research, while Rhino Health is recommended for managed infrastructure across institutions.
Since Jul 7Updated Nov 14, 2026Based on 52 AI answers
Reviewed by Dimitry Apollonsky
Brands AI mentions
Also mentioned rarely: Substra, Intel, Lifebit, Google Cloud, OpenFL, NVIDIA, PySyft, Fed-BioMed, Oracle, Openmined, Amazon Web Services, Duality Technologies, TensorFlow Federated, Siemens Healthineers, Apheris, Aridhia, NVFlare · nvidia.com, IBM, Intel SGX, NVIDIA Clara, Azure Machine Learning, Azure Confidential Computing, Hewlett Packard Enterprise, IBM Federated Learning, NVIDIA Inference (NVIDIA Inference Stack), Rhino Federated Computing, AMD, Google Cloud GKE Enterprise, Google Cloud Gemini Enterprise Agent Platform, Dell Technologies, FeatureCloud, FedML, Flip (formerly Flipgrid), Microsoft, NVIDIA DGX, Oracle Cloud Infrastructure Data Integration, Owkin Connect, MONAI, IBM Research, tracebloc, Vantage6, AWS Nitro Enclaves, AWS Outposts, Amazon HealthLake, AWS HealthOmics, Amazon SageMaker, AWS IoT Greengrass, AWS Auto Scaling, Azure Arc, BC Platforms, Beekeeper AI, Bitfount, Brane, Cancer AI Alliance, Google Cloud Healthcare API, NVIDIA CUDA Toolkit, DNAnexus, EGI (European Grid Infrastructure), Federated AI Technology Enabler, FedEMR.ai, Federated Learning Interoperability Platform, Taiwan Web Service Corporation, Flower Enterprise, Flywheel, Kaapana, FATE, Eli Lilly and Company, Linux Foundation, Merative, NVIDIA AI Enterprise, Neural Network Compression Framework, PetBioMit, PyTorch, Sherpa, Stand Together, Substra Foundation, Google Gemini, TensorOpera, Terraform, Lilly TuneLab
- NVFlareChatGPTGoogle AI Mode
- Microsoft AzureChatGPTGoogle AI Mode
- FlowerChatGPTGoogle AI Mode
- AWSChatGPTGoogle AI Mode
- Google Cloud PlatformChatGPTGoogle AI Mode
Questions and answers
ChatGPTNov 1, 2026
For a healthcare federated learning (FL) system, infrastructure is generally provided by a combination of cloud providers, federated learning platform vendors, and participating hospitals or healthcare institutions.
There isn't one company that provides the entire infrastructure. Several companies specialize in different layers.
1. Major infrastructure providers
These are the key companies and platforms to consider.
1. NVIDIA — NVIDIA FLARE
(Strong healthcare candidate)
- Federated learning orchestration.
- Secure communication between hospitals.
- Distributed model training and aggregation.
- Healthcare imaging and AI workflows.
Website:
2. Rhino Health
- Healthcare-specific federated computing.
- Connects hospitals and research institutions.
- Supports federated model training and deployment.
- Integrates with NVIDIA FLARE.
Website:
www.nvidia.comacademic.oup.comjiangxi.cs.uwm.eduwww.aicentre.co.ukwww.kcl.ac.uk
3. Flower Labs — Flower
- Open-source federated learning framework.
- Supports PyTorch, TensorFlow and other ML frameworks.
- Distributed training across multiple institutions.
- Flexible deployment across cloud and on-premises environments.
Website:
public.practice.gto.tech.gov.sgwww.aicentre.co.ukwww.kcl.ac.uk
4. Owkin — Substra
- Federated learning for healthcare and life sciences.
- Collaborative research across institutions.
- Data privacy and governance.
- Distributed model development.
Website:
academic.oup.comjiangxi.cs.uwm.edu
5. Intel — OpenFL
- Open-source federated learning framework.
- Distributed training infrastructure.
- Support for trusted execution environments.
- Suitable for multi-institution research.
Website:
academic.oup.comjiangxi.cs.uwm.edu
2. Cloud infrastructure providers
These companies provide the underlying computing infrastructure on which your federated learning system can run.
AWS
EC2, S3, EKS, networking, GPU infrastructure
Microsoft Azure
Virtual machines, Kubernetes, AI and security services
Google Cloud
Compute, Kubernetes, GPUs and AI infrastructure
AWS has documented a healthcare federated learning architecture using NVIDIA FLARE, with hospital-side clients accessing private medical imaging data and a server coordinating training.
3. Who provides what in your system?
A typical healthcare federated learning architecture looks like this:
Federated healthcare infrastructure
Hospital A
Patient data + local compute
Hospital B
Patient data + local compute
Hospital C
Patient data + local compute
Local model updates, not raw patient records
Federated learning platform
NVIDIA FLARE / Flower / OpenFL
Coordination, aggregation, security and orchestration
Cloud / on-premises infrastructure
AWS / Azure / Google Cloud / hospital data centers
Compute, storage, networking, GPUs and deployment
Global AI model
Aggregated model for evaluation and deployment at participating hospitals
4. My recommendation for your project
I would shortlist these combinations:
Architecture Infrastructure combination Open-source development Flower + AWS Healthcare-focused FL NVIDIA FLARE + AWS Managed healthcare platform Rhino Health + NVIDIA FLARE Research consortium OpenFL or Substra Flexible enterprise deployment Flower + Azure or AWS One particularly interesting real-world example is FLIP (Federated Learning Interoperability Platform), developed by the London AI Centre and deepc with Flower Labs, NVIDIA and AWS contributions. It is designed to support federated AI across healthcare institutions.
Website: FLIP open-source project.
5. One important question about your project
When you say you are building a federated learning system, are you planning to:
A. Build the entire infrastructure yourselves — connect hospitals, deploy local clients, build the aggregation server and manage security.
B. Use an existing infrastructure provider — such as NVIDIA FLARE, Rhino Health or Flower, and build your application on top.
C. Build a commercial healthcare FL platform — provide infrastructure as a service to hospitals, healthcare networks and pharmaceutical companies.
The distinction matters because it changes which companies you should partner with, which infrastructure you need to purchase, and where your actual product opportunity lies.
Pages cited in this answer18
- developer.nvidia.comNVIDIA FLARE | NVIDIA Developer
- aws.amazon.comPrivacy-Preserving Federated Learning on AWS with NVIDIA FLARE | AWS Partner Network (APN) Blog
- github.comGitHub - NVIDIA/NVFlare: NVIDIA Federated Learning Application Runtime Environment · GitHub
- rhinohealth.comRhino FC | Activate Sensitive Data for AI, Analytics, and Agents
- nvidia.comHealthcare Solutions from Partners & ISVs | NVIDIA
- academic.oup.comCODA: an open-source platform for federated analysis and machine learning on distributed healthcare data | Journal of the American Medical Informatics Association
- jiangxi.cs.uwm.eduSoftware Engineering Aspects of Federated Learning Libraries: A Comparative Survey
- aicentre.co.ukAI Centre and deepc Launch FLIP, an Open-Source Platform to Enable Health AI Innovation at Scale, in Partnerships with OneLondon and Flower Labs | AI Centre for Value Based healthcare
- kcl.ac.ukSecuring patient data in large-scale AI healthcare development: AI Centre and deepc launch FLIP | King's College London
- flower.aiFlower: Advance Collaborative Superintelligence
- public.practice.gto.tech.gov.sgAppendices | GTO Practices Portal
- owkin.comOwkin | Tackling the most pressing questions in medicine
- github.comGitHub - securefederatedai/openfederatedlearning: An Open Framework for Federated Learning. · GitHub
- aws.amazon.comCloud Computing Services - Amazon Web Services (AWS)
- azure.microsoft.comCloud Computing Services | Microsoft Azure
- cloud.google.comAI and Cloud Computing Services | Google Cloud
- aws.amazon.comPrivacy-Preserving Federated Learning on AWS with NVIDIA FLARE | AWS Partner Network (APN) Blog
- github.comGitHub - londonaicentre/FLIP: Federated Learning Interoperability Platform · GitHub
Google AI ModeNov 1, 2026
When building a federated learning (FL) system for healthcare, the infrastructure is rarely provided by a single vendor. Instead, it is typically assembled using a combination of open-source frameworks, specialized health-tech platforms , and enterprise cloud or hardware providers that allow hospitals to keep patient data secure behind their local firewalls.
The primary categories of infrastructure providers and tools break down as follows:
- Specialized Healthcare FL & Privacy Platforms: Companies like Lifebit (providing Trusted Research Environments and federated data genomics platforms) and Rhino Health specialize in federated computing and data orchestration specifically for clinical settings, allowing AI models to run where the data lives.
- Enterprise AI & Hardware Frameworks: NVIDIA FLARE (Federated Learning Application Runtime Environment) is heavily utilized across medical imaging and healthcare consortia as an enterprise-grade SDK/framework to coordinate federated workflows across hospital clusters.
- Open-Source Orchestration Frameworks: Many teams build their custom infrastructure on top of open-source projects like Flower, OpenFL (Intel Open Federated Learning), PySyft (OpenMined) , or FATE . These provide the underlying codebase for secure aggregation, differential privacy, and client-server communication.
- Cloud and MLOps Providers: Major cloud ecosystems (such as AWS, Google Cloud, and Microsoft Azure) or specialized infrastructure groups (like Taiwan Web Service Corporation) often host the central aggregation server or provide confidential computing environments, though the heavy lifting of local compute happens on-premise at the participating hospitals.
To help narrow down which infrastructure route fits your project, tell me:
- Are the participating hospitals on-premise , in the cloud , or a hybrid mix?
- What type of medical data are you focusing on (EHR/tabular, medical imaging, or genomics)?
- Are you looking for a fully managed enterprise platform or an open-source framework to build from scratch?
Google AI ModeNov 6, 2026
Several key companies and open-framework providersoffer the underlying orchestration software, infrastructure, and secure deployment environments needed to run federated learning (FL) systems across hospitals and health systems:- NVIDIA (NVIDIA FLARE): Provides an open-source, domain-agnostic federated learning SDK (
NVIDIA FLARE- NVIDIA FLARE Documentation ) widely adopted in healthcare and medical imaging research. It supports secure aggregation, workflow control, and integration with hospital high-performance computing (HPC) setups. - Flower Labs (Flower): Offers an open-source federated learning framework integrated with enterprise and clinical environments. For instance, Flower Labs partnered with Aridhia to host federated nodes directly inside hospital Digital Research Environments (DREs), significantly compressing multi-institution deployment times.
- Owkin: Specializes in AI and federated learning for drug discovery and oncology. Owkin builds decentralized research networks that allow pharmaceutical companies and academic medical centers to train machine learning models on patient cohorts without centralizing or exposing raw health records.
- Aridhia: Provides collaborative data and research workspace infrastructure (Digital Research Environments) that host federated learning client nodes directly within secure hospital IT perimeters, addressing compliance and data governance hurdles.
- IBM (IBM Research / Watson Health ecosystem): Has developed enterprise federated learning toolkits (such as IBM Federated Learning) designed to integrate with corporate and clinical cloud infrastructures while enforcing differential privacy and encryption protocols.
If you are looking at a specific use case, let me know:
- Are you focusing on medical imaging, genomics/drug discovery , or electronic health records (EHR)?
- Do you need an open-source framework to build yourself or a turnkey commercial platform?
ChatGPTNov 6, 2026
Companies offering the infrastructure for a healthcare federated learning (FL) system fall into several categories: federated learning platform providers, cloud infrastructure providers, healthcare data collaboration platforms, and privacy/security technology providers.
The important distinction is that federated learning typically does not require hospitals to transfer patient-level data into a central repository. Instead, infrastructure coordinates model training across participating hospitals, health systems, research institutions, or pharmaceutical companies while data remains under local control.
There is rarely one vendor that supplies the entire infrastructure. Most deployments combine several providers.
1. Leading federated learning infrastructure companies
These are the companies I would prioritize when evaluating the market.
1. Rhino Health
(Healthcare-focused platform)
- Provides the Rhino Federated Computing Platform (FCP).
- Supports distributed AI training across hospitals and research institutions.
- Includes federated orchestration, edge computing, data harmonization, and MLOps.
- Supports cloud and on-premises deployments.
Particularly relevant for multi-hospital medical imaging and clinical research.
2. NVIDIA
(Core technical infrastructure)
- Provides NVIDIA FLARE, an open-source federated learning framework.
- Provides GPU computing infrastructure for model training.
- Supports orchestration, distributed training, and secure communication.
- Can be deployed on-premises, in the cloud, or in hybrid environments.
Particularly attractive for computationally intensive medical imaging, genomics, and AI research.
3. Owkin
- Develops federated learning and collaborative AI technology for life sciences.
- Its Substra technology supports decentralized machine learning.
- Focuses on pharmaceutical research, drug discovery, and biomedical data collaboration.
- Has experience with multi-institution research initiatives.
Particularly relevant for pharmaceutical companies and academic medical centers.
4. Flower Labs
- Develops the Flower federated AI framework.
- Supports distributed machine learning across multiple organizations.
- Integrates with common ML frameworks such as PyTorch.
- Provides a flexible foundation for custom federated learning deployments.
Particularly relevant for organizations building their own FL infrastructure rather than purchasing a complete healthcare platform.
5. Intel
- Developed OpenFL, an open-source federated learning framework.
- Supports distributed training across institutions.
- Offers CPU and accelerator technologies for local training infrastructure.
- Relevant to organizations building customized, distributed research environments.
Particularly relevant for research institutions and organizations considering on-premises infrastructure.
6. Apheris
- Provides infrastructure for privacy-preserving data collaboration.
- Supports controlled computation across distributed datasets.
- Focuses on secure collaboration involving pharmaceutical and life sciences organizations.
- Relevant when data governance and permissions are as important as model training.
pmc.ncbi.nlm.nih.govwww.frontiersin.orgwww.meticulousresearch.com
2. Cloud infrastructure providers
These companies provide the underlying computing, storage, networking, security, and GPU resources needed to operate federated learning systems.
They are generally infrastructure partners rather than direct substitutes for specialized federated learning platforms.
Amazon Web Services (AWS)
- Cloud computing and GPU infrastructure.
- EC2, S3, EKS, networking, and security services.
- Supports deployments of NVIDIA FLARE and Rhino Health.
- Relevant to multi-hospital research networks.
aws.amazon.com.cdn.amazon.comwww.parse.glwww.meticulousresearch.com
Microsoft Azure
- Cloud computing, storage, GPU resources, and Kubernetes.
- Identity management and enterprise security.
- Supports hybrid and cloud-based healthcare AI architectures.
aws.amazon.com.cdn.amazon.comwww.parse.glwww.meticulousresearch.com
Google Cloud
- Cloud computing and GPU infrastructure.
- Kubernetes and distributed computing.
- Data engineering and AI/ML services.
- Potential foundation for custom federated learning deployments.
aws.amazon.com.cdn.amazon.comwww.parse.glwww.meticulousresearch.com
Oracle
- Cloud infrastructure and enterprise databases.
- Computing and storage for healthcare workloads.
- Relevant to organizations already operating Oracle healthcare environments.
www.nvidia.commedinform.jmir.orgsos-ch-dk-2.exo.iowww.meticulousresearch.compmc.ncbi.nlm.nih.govwww.frontiersin.orgwww.parse.gl
3. Additional healthcare data and privacy infrastructure companies
These companies are worth investigating because federated learning requires more than distributed model training. Healthcare organizations also need data harmonization, governance, privacy-preserving computation, and interoperability.
Company Infrastructure capability Potential role DNAnexus Cloud-based genomic and biomedical data infrastructure Genomics and precision medicine BC Platforms Biomedical data management and research infrastructure Multi-institution clinical and genomic research Lifebit Federated biomedical data analysis and secure research environments Population-scale biomedical research Duality Technologies Privacy-enhancing computation and encryption Secure collaboration on sensitive data Dell Technologies Enterprise servers, storage, and on-premises infrastructure Hospital-based computing Hewlett Packard Enterprise Enterprise computing, HPC, and hybrid cloud Distributed compute infrastructure Siemens Healthineers Medical imaging and healthcare technology infrastructure Imaging-based AI collaboration These companies occupy different parts of the infrastructure stack; not all offer a dedicated federated learning platform. The healthcare research platform landscape also identifies DNAnexus, BC Platforms, and several dedicated FL providers as distinct categories of infrastructure.
4. How these companies fit together
A typical healthcare federated learning architecture could look like this:
Healthcare federated learning architecture
Federated learning coordination layer
Rhino Health · NVIDIA FLARE · Flower · OpenFL · Owkin/Substra
Coordinates training, aggregation, and distributed workflows
Hospital A
EHR, imaging, local compute
Hospital B
EHR, imaging, local compute
Hospital C
EHR, imaging, local compute
Raw patient-level data stays within each participating institution in the intended architecture.
Underlying compute and infrastructure
AWS · Microsoft Azure · Google Cloud · Dell · HPE · NVIDIA
Compute, GPU acceleration, storage, networking, and deployment environments
Privacy, security, and data governance
Apheris · Duality · cloud security services · institutional security systems
Access controls, encryption, auditability, and privacy-enhancing technologies
One important technical consideration: federated learning does not automatically guarantee patient privacy or HIPAA compliance. Model updates can still expose sensitive information. Secure aggregation, differential privacy, encryption, access controls, and appropriate governance may be needed.
5. My recommended shortlist
If I were evaluating companies to build or procure a healthcare federated learning infrastructure, I would prioritize them as follows:
Priority Company Why investigate 1 Rhino Health Healthcare-specific federated computing platform 2 NVIDIA Federated learning framework and GPU infrastructure 3 AWS Cloud infrastructure and deployment ecosystem 4 Owkin Pharmaceutical and biomedical research collaboration 5 Flower Labs Flexible federated learning framework 6 Apheris Privacy-preserving healthcare and life sciences data collaboration 7 Intel Open-source federated learning and compute infrastructure 8 Microsoft Azure Enterprise cloud and hybrid infrastructure A particularly interesting example is the combination of Rhino Health and AWS. AWS has documented how Rhino's platform supports cross-organizational healthcare AI collaboration, including deployments involving pharmaceutical companies.
6. Important distinction: three different markets
For competitive analysis or market mapping, I would separate these companies into three groups:
- Federated learning infrastructure vendors: Rhino Health, NVIDIA, Flower Labs, Intel, and Owkin.
- Underlying infrastructure vendors: AWS, Microsoft, Google, Oracle, Dell, and HPE.
- Healthcare data collaboration and privacy vendors: Apheris, Duality Technologies, DNAnexus, BC Platforms, and Lifebit.
This distinction matters because a company selling GPUs or cloud storage is not necessarily competing with a company selling a federated learning orchestration platform.
My overall assessment: Rhino Health, NVIDIA, Owkin, and Apheris are especially interesting starting points if your focus is the specialized healthcare federated learning infrastructure market. AWS, Azure, and Google Cloud are important complementary infrastructure providers.
7. Further research
One question would help narrow this competitive landscape considerably:
Are you primarily interested in companies that:
- A. Sell the actual federated learning software/platform to hospitals and health systems.
- B. Supply the underlying hardware, cloud computing, storage, and networking infrastructure.
- C. Provide an end-to-end solution, including software, infrastructure, security, and healthcare data integration.
I can develop a more detailed competitive landscape with company profiles, products, business models, healthcare customers, and partnerships for whichever segment you are evaluating.
Pages cited in this answer18
- nvidia.comHealthcare Solutions from Partners & ISVs | NVIDIA
- aws.amazon.com.cdn.amazon.comCollaborative AI Model Training with Rhino Federated Computing on AWS
- parse.gl · Parse's own pageHealthcare federated learning infrastructure in AI Compute Infrastructure Solutions | Parse
- rhinohealth.comRhino FC | Activate Sensitive Data for AI, Analytics, and Agents
- medinform.jmir.orgJMIR MEDICAL INFORMATICS
- sos-ch-dk-2.exo.ioUPM_HBP_MIP_FINAL_DT
- meticulousresearch.comFederated Learning Outlook 2026–2034 / Meticulous Next™ | Meticulous Research
- developer.nvidia.comNVIDIA FLARE | NVIDIA Developer
- owkin.comOwkin | Tackling the most pressing questions in medicine
- pmc.ncbi.nlm.nih.govSelecting medical research data platforms for translational biomedical research: a five-tier overview and requirement-weighted assessment framework - PMC
- frontiersin.orgSelecting medical research data platforms for translational biomedical research: a five-tier overview and requirement-weighted assessment framework
- flower.aiFlower: Advance Collaborative Superintelligence
- github.comGitHub - securefederatedai/openfederatedlearning: An Open Framework for Federated Learning. · GitHub
- apheris.comSuperior drug discovery models | Apheris
- aws.amazon.comAWS for Healthcare & Life Sciences | AWS
- azure.microsoft.comAzure healthcare
- cloud.google.comCloud for Healthcare and Life Sciences (HCLS) | Google Cloud
- oracle.comOracle Health
- NVIDIA (NVIDIA FLARE): Provides an open-source, domain-agnostic federated learning SDK (
Also asked as
- I need to find providers for healthcare federated learning infrastructure. Who should I look at?