Graph databases & knowledge graphs · Amazon Web Services

Amazon Neptune

Fully managed AWS graph database supporting both property-graph (Gremlin, openCypher) and RDF (SPARQL) query models.

Visit Amazon Neptune Alternatives

Amazon Neptune is a fully managed graph database with no self-hosted option: AWS operates the cluster, and you choose at creation whether it behaves as a property graph (queried with Apache TinkerPop Gremlin or openCypher) or an RDF triple store (queried with SPARQL) — one of the few engines that supports both models on the same infrastructure. Neptune Serverless scales capacity automatically in Neptune Capacity Units and bills per second. Neptune Analytics is a separate in-memory engine for running graph algorithms (PageRank, community detection, similarity) and vector search over data loaded from Neptune Database or Amazon S3. It integrates natively with IAM for access control, S3 for bulk loading, and SageMaker for Neptune ML, which trains graph neural networks directly on the stored graph.

At a glance

Vendor Amazon Web Services
Pricing model Usage-based
Free tier No
Deployment Cloud
Open source No
Best for AWS-centric teams who want a managed graph database without operating clusters, with flexibility to pick property-graph or RDF.

Pricing

Billed per instance-hour plus storage and I/O for provisioned clusters, or per Neptune Capacity Unit for Serverless.

Plan Price Notes
Neptune Database (Standard, db.r5.large) $0.348/hour example on-demand instance rate; varies by instance type and region
Neptune Database Storage $0.10/GB-month Standard configuration; I/O-Optimized storage is $0.225/GB-month with no separate I/O charge
Neptune Database I/O (Standard) $0.20 per million requests not charged on I/O-Optimized configuration
Neptune Analytics $7.68 per m-NCU-hour paused graphs billed at 10% of normal compute price

Prices read from the vendor's own page on September 21, 2026. Vendors change prices; check the source before you budget.

Features

  • Dual model: property graph (Gremlin, openCypher) or RDF (SPARQL)
  • Neptune Serverless with per-second capacity billing
  • Neptune Analytics for algorithms and vector search
  • Neptune ML for graph neural network predictions via SageMaker
  • Multi-AZ high availability with automated failover
  • Bulk loader from Amazon S3
  • IAM-based fine-grained access control
  • Automated, incremental backups

Integrations

Profile last reviewed September 21, 2026

Head to head

Alternatives

Neo4j Native property-graph database using Cypher, the language behind the new ISO GQL standard, for relationship-heavy analytics. JanusGraph Open-source distributed property-graph database queried with Gremlin, with pluggable storage and indexing backends. TigerGraph Massively parallel native graph database with its own GSQL language, built for deep multi-hop analytics at scale. Stardog Enterprise RDF knowledge-graph platform using SPARQL, with OWL/RDFS reasoning and federation over relational sources.

Amazon Neptune in the index now

Terms to know

Data modeling The practice of organizing data into entities, relationships and schemas that make it accurate and efficient to query. Machine learning Building models that learn patterns from data to make predictions or decisions, rather than following hand-coded rules. Neural network A model made of layers of connected nodes that learns to map inputs to outputs by adjusting connection weights during training. Semantic search Search that matches results to a query's meaning rather than its exact keywords, using vector similarity between embeddings. Community detection Identifying groups of nodes in a network that are more densely connected to each other than to the rest of the graph. PageRank An algorithm that scores a node's importance by the number and importance of other nodes linking to it.

Related guides

Graph and network analytics, explained How relationships between things become data — the algorithms that find structure in a network and where graph databases fit. How to choose a graph database Decide property graph vs RDF first, then managed vs self-hosted and how the engine scales — the query language follows from those, not the other way round.