Neptune vs DocumentDB Pricing: Cost Comparison and When Each Fits

Neptune (graph) and DocumentDB (MongoDB-compatible document store) are AWS-managed alternatives to running open-source graph databases or MongoDB on EC2. Both compete with self-managed deployments, MongoDB Atlas, and specialised vendors. The cost model is similar to Aurora: instance-hours, storage-consumed, I/O requests, and backup storage. The optimisation moves and negotiation levers track the broader database stack but with specific traps worth knowing.

The shared cost model

Both services use the Aurora-style cluster architecture (one writer plus up to 15 readers, shared storage volume). Cost optimisation patterns mirror Aurora: instance right-sizing, replica strategy, I/O hygiene.

Neptune pricing detail

Neptune instance pricing tracks the Aurora-equivalent instance class. A db.r6g.xlarge Neptune cluster bills:

• ~$0.348 per hour per instance (us-east-1 on-demand).
• $0.10 per GB-month storage consumed.
• $0.20 per million I/O requests.
• Multi-AZ failover via additional reader instances.

Neptune Serverless prices in Neptune Capacity Units (NCUs) per second. Configuration mirrors Aurora Serverless v2: minimum and maximum NCU bounds, with the workgroup scaling between them.

The cost optimisation surface for Neptune is small:

• Right-size aggressively; graph query patterns rarely justify large instance classes outside specific traversal workloads.
• Use Serverless for development and ad-hoc analytical work.
• Audit I/O patterns; complex graph traversals can generate disproportionate I/O.
• Reserved instances available on 1- and 3-year terms with standard Aurora-equivalent discounts (~30 to 65 percent).

DocumentDB pricing detail

DocumentDB instance pricing similarly tracks Aurora-equivalent classes. A db.r6g.large cluster bills approximately $0.277 per hour per instance plus storage and I/O.

DocumentDB Elastic (introduced 2023) adds horizontal sharding for workloads exceeding single-cluster capacity, with different pricing:

• Per shard instance hour rate.
• Storage and I/O still consumed-based.
• Useful for very large document estates (multi-TB).

The optimisation moves:

• Standard Aurora-style right-sizing.
• Reserved instance coverage on steady-state.
• Cross-region replication only where justified by latency or compliance.
• I/O audit on write-heavy collections.

The MongoDB Atlas alternative

MongoDB Atlas runs on AWS infrastructure but is sold through MongoDB Inc., not AWS. The cost comparison:

• Headline pricing. Atlas dedicated clusters tend to be 15 to 35 percent more expensive than equivalent DocumentDB capacity at list price.
• Feature parity. Atlas supports the full MongoDB feature set; DocumentDB has a compatibility layer with documented gaps (transactions, certain aggregation operators, change streams behaviour).
• EDP routing. MongoDB Atlas can be purchased through AWS Marketplace, routing the spend through your AWS EDP for commitment drawdown. This makes the headline price difference economically irrelevant for most buyers. See ISV workloads through EDP for the routing mechanics.
• Negotiation surface. MongoDB direct negotiation is typically softer than AWS direct. Marketplace private offers from MongoDB regularly come in 15 to 25 percent below Atlas list.

For enterprises with sustained MongoDB-style workloads above $30K/month, the Atlas-via-Marketplace path frequently beats DocumentDB on total cost once EDP drawdown is counted.

The self-managed alternative

Self-managed MongoDB or Neo4j on EC2 with disciplined operations:

• Direct EC2 instance cost (with Savings Plan coverage) is roughly 30 to 50 percent below DocumentDB / Neptune instance cost.
• Operational overhead is substantial: backups, patching, replica set management, disaster recovery.
• Justified at very large scale where the operational team already has the expertise.
• Not appropriate for teams without dedicated database engineering capacity.

The EDP private pricing layer

Neptune and DocumentDB roll into the broader database scope of EDP negotiation:

• Standard EDP tier discounts on instance pricing.
• Custom storage and I/O concessions at multi-TB / billion-request scale.
• Cross-region replication transfer discounts.
• Reserved instance custom discounting on sustained spend above $50K/month.

The strategic question at EDP renewal is not just price-per-instance but whether the workload should be on DocumentDB / Neptune at all or routed through Marketplace to a specialised vendor.

Case study: $480K DocumentDB / Neptune estate

A media company running:

• DocumentDB cluster (3 instances on db.r5.2xlarge) for content metadata. $290K annualised.
• Neptune cluster (2 instances on db.r5.xlarge) for content recommendation graph. $190K annualised.

Audit findings:

• DocumentDB instance utilisation averaged 38 percent CPU. Right-sizing candidate.
• Neptune instances oversized for the actual traversal pattern.
• Zero reserved instance coverage on either cluster.
• MongoDB Atlas private offer at 22 percent below DocumentDB equivalent capacity available through Marketplace.

Interventions:

• Right-sized DocumentDB to db.r6g.large x 3. $98K annualised saving.
• Right-sized Neptune to db.r6g.large x 2 plus Serverless secondary for ad-hoc analytics. $72K annualised saving.
• 3-year reserved instance coverage on stable DocumentDB instances. $46K annualised saving.
• Modelled MongoDB Atlas via Marketplace as alternative; deferred migration but baselined the option for EDP renewal leverage.

Net first-year saving: $216K, a 45 percent reduction without migrating to Atlas. Atlas option preserved as renewal leverage.

Action checklist

1. Inventory Neptune and DocumentDB clusters. Measure utilisation and I/O patterns.
2. Right-size instances; consider Graviton (r6g) where supported.
3. Apply reserved instance coverage on stable clusters.
4. For DocumentDB at scale, model MongoDB Atlas via Marketplace as alternative.
5. For Neptune, model Neo4j on EC2 only if operational capacity exists.
6. Scope Neptune / DocumentDB spend into EDP negotiation.
7. Use Marketplace alternative as renewal leverage even if not migrating.
8. Contact our advisory team for a graph / document database cost audit benchmarked against $2.4B+ of reviewed AWS spend.

Neptune and DocumentDB occupy a narrow segment of the AWS database stack where the alternative-platform question is as important as the in-service optimisation question. See our AWS database cost strategy guide for the broader context.