FHIR storage budgets almost always underestimate the bill. The number you commit to on a slide deck is a projection of current-state resources at their compressed size, and the number you see on the invoice a year later includes history rows, attachments, indexes, backups, and audit trails you never sized for. The gap is not small; it is often a factor of three to five.
Naming the hidden costs in advance is the discipline that turns storage surprises into scheduled expenses. For related walkthroughs, our FHIR explainer library collects the surrounding material.
The Line Items Most Plans Miss
A storage budget that lists only current-state resource size will drift immediately. The realistic budget names each of these:
- Current-state resources in their serialized form.
- Version-history rows appended on every write.
- Search indexes on every SearchParameter the server supports.
- Attachments carried inline through Observation.valueAttachment or DocumentReference.
- Audit trail entries recording every access.
- Backups retained for the disaster-recovery horizon.
Every one of these grows independently. A pass through the site's FHIR storage bill projector multiplies through the per-engine footprint so you can see the total against a patient count instead of guessing.
Version History Is the Sneaky One
Every update to a FHIR resource appends a new row to the history table. A Patient that gets updated five times contributes five rows. Multiply across every resource type and every regular update rhythm and the history table quickly reaches parity with the current-state table.
At year three, most FHIR deployments carry a history table that is two to three times the size of current state. Plans that only budget for current state miss this entirely. For the specific split, history table vs current state: doubling your storage bill quietly covers the mechanics.
Attachments Ambush Everyone Eventually
Attachments in FHIR are the storage line that shows up on the invoice unexpectedly. A single DocumentReference with an inline PDF can carry a payload that is fifty times the size of an average Observation. A workflow that leans on inline attachments will grow storage at a rate that raw resource counts never predict.
The disciplined pattern is to store attachments on object storage and reference them from FHIR resources. That trade is usually the difference between a manageable bill and an unmanageable one. For the specific ambush shape, attachment storage in FHIR: when Observation.valueAttachment ambushes you covers the pattern.
Indexes Cost Real Space
Every SearchParameter the server supports has a corresponding index. Wide search-parameter coverage is comfortable at query time and expensive at storage time. A deployment that keeps every possible SearchParameter enabled can spend twenty to thirty percent of storage on indexes alone.
Plans should name the indexes explicitly. The exercise of listing them tends to expose SearchParameters that are enabled but never queried, which is a straight-line cost saving. For the resource-level footprint that indexes multiply against, per-resource storage footprint: what actually takes space covers the base sizes.
Backups Compound Everything
Every backup snapshot preserves the current storage state. A daily backup retention window of ninety days means the deployment is paying for roughly ninety copies of the archive at any moment. Compression softens the number but not by an order of magnitude.
Backup retention is one of the highest-leverage cost knobs. A tiered backup strategy that keeps hourly for a day, daily for a week, weekly for a month, and monthly for a year absorbs the actual audit needs at a fraction of the naive cost.
The Retention Policy Interaction
Retention policies interact with every line item above. Extending retention doubles the archive line without changing anything else. Tiering the retention across storage tiers softens the cost without giving up the data. For the retention framing specifically, how retention policies shape long-term FHIR storage cost walks through the trade.
Every FHIR deployment eventually has a storage conversation with its finance team. Naming the hidden costs before that conversation is the shortest path to defending the plan.
Sources
- HL7 FHIR core specification of resource lifecycle including - HL7 FHIR core specification of resource lifecycle including versioning that shapes hidden storage cost