The deployment model for the CMS-0057-F FHIR platform is one of the choices that gets less attention than vendor selection but has substantial operational and cost consequences. Cloud-hosted (managed services from Smile, 1upHealth, Microsoft, AWS, Google) shifts operational burden to the vendor. Self-hosted (Smile CDR on-prem, HAPI FHIR, InterSystems IRIS, Edifecs) keeps operational control with the payer. Both are conformant; the choice has different implications. For the FHIR architecture knowledge base on this site, this is the deployment-model decision.
What Cloud-Hosted Actually Includes
A cloud-hosted FHIR platform runs in the vendor's (or cloud provider's) infrastructure. The vendor handles server provisioning, scaling, patching, monitoring, backup, security infrastructure, and FHIR-engine tuning. The payer integrates against the FHIR API surface and configures use-case-specific behavior. Operational issues that come from infrastructure are the vendor's responsibility.
The pricing model varies. Some vendors price for the platform as a whole; others price per resource, per API call, or per member. Cloud-hosted is typically more expensive at the platform line than the equivalent self-hosted license, but the total cost of ownership often comes out comparable or favorable once operational headcount is counted.
What Self-Hosted Actually Requires
A self-hosted FHIR platform runs on infrastructure the payer manages. The payer team handles deployment, scaling, patching, monitoring, backup, security infrastructure, and FHIR-engine tuning. The vendor (if there is one) handles licensing and support; operations are the payer's.
The headcount for production-grade FHIR operations is substantial. A payer running Smile CDR or InterSystems IRIS on-premise at production scale typically needs three to six engineers handling the platform alone (excluding the higher-level integration work).
Where Cloud-Hosted Wins
Cloud-hosted wins on operational simplicity. The payer team focuses on integration and CMS-0057-F-specific work; the infrastructure work happens at the vendor. Operational scaling (more API traffic post-2027) happens transparently; the payer does not have to provision additional capacity manually.
Cloud-hosted also wins on time-to-deployment. The 3-6 month go-live timelines that mid-market payers need to hit January 1, 2027, are realistic with cloud-hosted offerings. Self-hosted deployments typically take longer because the infrastructure work has to happen alongside the integration work.
Cloud-hosted wins on IG maintenance for the FHIR engine itself. The vendor maintains FHIR conformance across IG updates as part of the managed offering. Self-hosted payers may need to upgrade their infrastructure to track IG changes.
Where Self-Hosted Wins
Self-hosted wins on data control. The data lives in payer-controlled infrastructure. Regulatory and contractual requirements that mandate data residency in payer-owned environments are met without exception language. State Medicaid contracts often include such requirements.
Self-hosted wins on operational customization. The payer can tune the platform for specific use cases, integrate with internal systems at the infrastructure level, and run experimental configurations. Cloud-hosted managed services rarely allow this depth of customization.
Self-hosted wins on long-term cost predictability for payers with very high API traffic. Cloud-hosted pricing that scales with API volume can become more expensive than self-hosted at very high scale. The break-even point varies by vendor and traffic profile.
The Hybrid Pattern That Some Large Payers Run
Some large payers run hybrid deployments. The Patient Access API (highest volume, most public) runs cloud-hosted for the scaling and operational simplicity. The Prior Authorization API and the back-office UM integration run on-premise for the data-control and operational-customization benefits.
The hybrid is operationally complex but fits payers with clear use-case-specific deployment preferences. The integration layer between cloud-hosted and on-premise pieces needs careful design.
How to Pick Honestly
The decision usually comes down to four factors. Regulatory and contractual requirements (do they mandate data residency or specific hosting). Operational maturity (does the payer have strong existing infrastructure operations). Timeline (is 3-6 month go-live a constraint). API traffic projection (will post-2027 traffic stay within cloud-hosted economic ranges or grow beyond them).
For most mid-market payers in 2026, cloud-hosted is the default unless specific requirements push otherwise. For specific cloud-native options, the Top 5 cloud-native FHIR hosting options for US health payers covers the leaders. For self-hosted on-premise options, the Best on-premise FHIR platforms for regulated payer deployments covers the alternatives.