Is MemoryDB more expensive than ElastiCache?

Yes. For the same node type, MemoryDB's per-node rate is higher than ElastiCache, and MemoryDB adds a per-GB data-written charge (~$0.20/GB) that ElastiCache doesn't have. The premium pays for a durable multi-AZ transaction log that lets MemoryDB serve as a primary database, not just a cache.

What is the MemoryDB data-written charge?

Every GB written to MemoryDB's durable transaction log bills at roughly $0.20/GB. ElastiCache has no equivalent because it isn't durable. For write-heavy workloads this charge can rival the node cost, so it's the line to model carefully when migrating from ElastiCache.

When should I use MemoryDB instead of ElastiCache?

Use MemoryDB when you need Redis to be the system of record, durable across failures without a separate database behind it. Use ElastiCache when Redis is a cache in front of another datastore and you can rebuild it on failure. The durability requirement, not performance, is the deciding factor.

How do I count nodes for either service?

For cluster-mode ElastiCache replication groups and MemoryDB, total nodes are shards multiplied by (1 + replicas per shard). A 3-shard cluster with 1 replica each is 6 nodes, each billed at the node-type hourly rate. Both services price node-hours the same way.

How does C3X estimate the difference?

C3X prices node-hours for both from node type and the shard/replica counts, and adds MemoryDB's data-written and snapshot dimensions from your usage file. Putting both resources in a Terraform plan and running c3x estimate shows the side-by-side cost before you commit.

ElastiCache vs MemoryDB cost: when durability is worth paying for

Quick answer

MemoryDB costs more than ElastiCache for the same node type and adds a ~$0.20/GB data-written charge ElastiCache doesn't have. You're paying for a durable multi-AZ log that makes MemoryDB a primary datastore. If Redis is a cache in front of another database, ElastiCache is cheaper and correct. If Redis is your system of record, MemoryDB's premium buys the durability you'd otherwise build yourself.

Both run Redis-compatible in-memory workloads on AWS, and at a glance they look interchangeable. They are not, and the cost difference traces directly to one feature: durability.

The pricing difference

ElastiCache (cache)

Node-hours by node type (e.g. cache.r6g.large ~$0.204/hour)
Backup storage beyond one free snapshot per node
No charge for writes

MemoryDB (durable datastore)

Node-hours by node type, at a higher rate than the equivalent ElastiCache node
Data written to the durable log: ~$0.20/GB
Snapshot storage beyond the free allowance

Node count works the same for both: shards times (1 + replicas per shard). The divergence is the per-GB data-written charge, which is unique to MemoryDB and can be the larger line for write-heavy workloads.

What the premium buys

MemoryDB keeps a multi-AZ transaction log, so committed writes survive a node or zone failure. That lets it be the primary store for data you can't afford to lose, no separate database required. ElastiCache is an in-memory cache: fast, but if a node fails you rebuild from the source of truth. The price gap is the cost of not needing that source of truth.

When to choose which

ElastiCache: Redis sits in front of a database (RDS, DynamoDB) and can be rebuilt on failure. Most caching use cases.
MemoryDB: Redis is the system of record, session stores you can't lose, leaderboards, real-time state with durability needs.

See the real number before you decide

The data-written charge makes MemoryDB hard to estimate by intuition. Price both directly: aws_elasticache_replication_group and aws_memorydb_cluster both have full pricing breakdowns, and C3X estimates either from your Terraform with the write volume from a usage file.

# Compare both in one estimate
c3x estimate --path . --usage c3x-usage.yml

For the broader caching picture, see DynamoDB on-demand vs provisioned for the durable-store alternative.