← Databases
Graph Database

Amazon Neptune Pricing

Amazon Neptune is a fast, reliable, fully managed graph database service optimized for billions of relationships and millisecond traversals. Supports openCypher, Apache TinkerPop Gremlin, and W3C SPARQL with flexible Neptune Serverless (1–128 NCUs), provisioned instances, and in-memory Neptune Analytics.

34AWS Regions
3 ModelsopenCypher · Gremlin · SPARQL
1–128 NCUServerless Dynamic Range
128 TiBMax Distributed Storage

Deployment Options: Serverless, Provisioned & Analytics

Deploy Neptune according to your workload demands — auto-scaling Serverless for spiky traffic, provisioned fleets for steady state, or Neptune Analytics for massive algorithmic computations.

Auto-Scaling · Pay as you go

Neptune Serverless

Instantly and automatically adjusts compute capacity in increments of 0.5 Neptune Capacity Units (NCUs) up to 128 NCUs. Eliminates manual capacity provisioning.

  • Base Pricing: $0.120 per NCU-Hour
  • Memory Density: ~2 GiB RAM per NCU
  • Scaling Model: Automated instant scaling (0.5 to 128 NCUs)
  • Best For: Variable query traffic, dev/test environments, and unpredictable graph workloads.
Dedicated Instance Fleet

Neptune Provisioned

Dedicated compute instances across AWS Graviton and Intel Xeon families with up to 15 read replicas sharing cluster storage. Predictable throughput for steady OLTP traffic.

  • Base Pricing: From $0.086 / instance-hour
  • Memory Density: Up to 512 GiB RAM per node
  • Scaling Model: Manual instance resizing & read replica scaling
  • Best For: Steady, high-throughput OLTP graph query applications with predictable loads.
In-Memory Graph Engine

Neptune Analytics

Purpose-built, memory-optimized graph analytics engine with integrated vector search. Runs complex graph algorithms (PageRank, clustering, shortest path) in seconds on tens of billions of edges.

  • Base Pricing: $0.033 per m-NCU Hour
  • Memory Density: Memory-optimized graph units (m-NCU)
  • Scaling Model: On-demand graph capacity allocation
  • Best For: Global graph algorithms, fraud ring detection, vector search, and batch analytics.

Storage Classes: Standard vs. I/O-Optimized

Neptune decouples compute from storage with 6-way replication across 3 AZs. Select your storage configuration based on graph traversal intensity.

Standard Storage

Ideal for graph workloads with light-to-moderate query traversals where storage volume exceeds I/O request costs.

$0.100/ GB-month

+ $0.200 per 1,000,000 read/write I/O requests

Recommended for Deep Traversals

I/O-Optimized Storage

Designed for I/O-intensive graph traversals, neighborhood searches, and frequent edge traversals. Eliminates all I/O billing.

$0.300/ GB-month

$0.000 (Included) zero I/O request fees

Provisioned Compute Instances (US East, N. Virginia)

On-demand database instances for primary writer and read replica nodes. Graviton3 instances deliver the highest memory bandwidth for relationship scans.

Instance TypeFamilyArchitecturevCPURAMHourly RateMonthly Est. (730h)
db.r7g.largeMemory Optimizedarm64 (Graviton3)216 GiB$0.359/hr$262.07/mo
db.r7g.xlargeMemory Optimizedarm64 (Graviton3)432 GiB$0.718/hr$524.14/mo
db.r7g.2xlargeMemory Optimizedarm64 (Graviton3)864 GiB$1.436/hr$1048.28/mo
db.r7g.4xlargeMemory Optimizedarm64 (Graviton3)16128 GiB$2.872/hr$2096.56/mo
db.r7g.8xlargeMemory Optimizedarm64 (Graviton3)32256 GiB$5.744/hr$4193.12/mo
db.r7g.12xlargeMemory Optimizedarm64 (Graviton3)48384 GiB$8.616/hr$6289.68/mo
db.r7g.16xlargeMemory Optimizedarm64 (Graviton3)64512 GiB$11.488/hr$8386.24/mo
db.r6g.largeMemory Optimizedarm64 (Graviton2)216 GiB$0.348/hr$254.04/mo
db.r6g.xlargeMemory Optimizedarm64 (Graviton2)432 GiB$0.696/hr$508.08/mo
db.r6g.2xlargeMemory Optimizedarm64 (Graviton2)864 GiB$1.392/hr$1016.16/mo
db.r6g.4xlargeMemory Optimizedarm64 (Graviton2)16128 GiB$2.784/hr$2032.32/mo
db.r6g.8xlargeMemory Optimizedarm64 (Graviton2)32256 GiB$5.568/hr$4064.64/mo
db.r6g.12xlargeMemory Optimizedarm64 (Graviton2)48384 GiB$8.352/hr$6096.96/mo
db.r6g.16xlargeMemory Optimizedarm64 (Graviton2)64512 GiB$11.136/hr$8129.28/mo
db.r6i.largeMemory Optimizedx86_64 (Intel)216 GiB$0.380/hr$277.40/mo
db.r6i.xlargeMemory Optimizedx86_64 (Intel)432 GiB$0.760/hr$554.80/mo
db.r6i.2xlargeMemory Optimizedx86_64 (Intel)864 GiB$1.520/hr$1109.60/mo
db.r6i.4xlargeMemory Optimizedx86_64 (Intel)16128 GiB$3.040/hr$2219.20/mo
db.r6i.8xlargeMemory Optimizedx86_64 (Intel)32256 GiB$6.080/hr$4438.40/mo
db.r6i.12xlargeMemory Optimizedx86_64 (Intel)48384 GiB$9.120/hr$6657.60/mo
db.r6i.16xlargeMemory Optimizedx86_64 (Intel)64512 GiB$12.160/hr$8876.80/mo
db.t4g.mediumBurstablearm64 (Graviton2)24 GiB$0.086/hr$62.78/mo
db.t3.mediumBurstablex86_64 (Intel)24 GiB$0.108/hr$78.84/mo

Global Regional Rates Comparison

Compare Neptune Serverless NCU, standard storage, and compute instance costs across global AWS regions.

Region CodeRegion NameServerless NCUStandard StorageI/O-Optimizeddb.r6g.large Rate
us-east-1US East (N. Virginia)$0.120/NCU-hr$0.100/GB-mo$0.300/GB-mo$0.348/hr
us-east-2US East (Ohio)$0.120/NCU-hr$0.100/GB-mo$0.300/GB-mo$0.348/hr
us-west-2US West (Oregon)$0.120/NCU-hr$0.100/GB-mo$0.300/GB-mo$0.348/hr
eu-west-1Europe (Ireland)$0.132/NCU-hr$0.110/GB-mo$0.330/GB-mo$0.383/hr
eu-central-1Europe (Frankfurt)$0.143/NCU-hr$0.119/GB-mo$0.357/GB-mo$0.414/hr
ap-southeast-1Asia Pacific (Singapore)$0.132/NCU-hr$0.110/GB-mo$0.330/GB-mo$0.398/hr
ap-northeast-1Asia Pacific (Tokyo)$0.144/NCU-hr$0.120/GB-mo$0.360/GB-mo$0.418/hr
sa-east-1South America (São Paulo)$0.168/NCU-hr$0.140/GB-mo$0.420/GB-mo$0.505/hr

Neptune Cost Optimization Playbook

1. Guardrail Serverless NCU Ranges

Configure minimum and maximum Neptune Capacity Units (NCUs) on Neptune Serverless. Setting a ceiling (e.g. 16 or 32 NCUs) protects your budget against rogue deep graph traversals or recursive queries that trigger aggressive scaling.

2. The I/O-Optimized Tipping Point

Graph queries with high branching factors (e.g. 3+ hop traversals) generate huge amounts of read I/Os. If I/O request fees exceed 25% of your total storage spend, enabling I/O-Optimized storage directly reduces your invoice.

3. Separate OLAP with Neptune Analytics

Running global graph algorithms (PageRank, Louvain community detection) on an OLTP Neptune Database requires giant instances. Use ephemeral Neptune Analytics graphs (at $0.033/m-NCU hr) to run jobs in seconds, then shut them down.

4. Leverage Graviton3 (R7g) Instances

Graph query performance is heavily memory-bandwidth bound. Graviton3 (db.r7g) instances offer up to 50% more DDR5 memory bandwidth than Graviton2, enabling significantly faster edge traversals at virtually identical cost.