AWS ElastiCache Serverless cost: pay for data stored and compute used

Quick answer

ElastiCache Serverless bills per GB-hour of data stored and per ElastiCache Processing Unit (ECPU) consumed by reads and writes, scaling automatically with no nodes to provision. It suits spiky or unpredictable caches where node-based provisioning would waste capacity; for large, steady caches, provisioned nodes can be cheaper. Match the model to the traffic shape.

ElastiCache Serverless removes node sizing from the caching decision. Instead of provisioning instances, you pay for the data the cache holds and the compute your operations consume, and it scales up and down automatically. That convenience is priced, so the question is whether your workload benefits from it.

Two usage dimensions

Dimension	Bills for
Data stored	Per GB-hour retained in the cache
ECPUs	Compute for reads and writes

Data stored scales with your working set, and ECPUs scale with how many operations you run and how large they are. A cache that is small and lightly used costs little; a large, hot cache consumes both dimensions heavily.

Serverless vs provisioned nodes

Serverless wins when traffic is spiky, unpredictable, or often low, because you do not pay for a large provisioned node sitting mostly idle, and there is a minimum floor rather than a full node. For a large cache under steady, heavy load, a right-sized provisioned node (especially with a reservation) can be cheaper per unit than the Serverless dimensions. It is the same steady-versus-spiky tradeoff as everywhere else.

Controlling Serverless cost

Keep the working set lean so data-stored charges stay low, use sensible TTLs so the cache does not retain data it will not reuse, and make operations efficient to reduce ECPU consumption. And remember the cache should deflect more downstream cost than it consumes, the caching break-even that justifies any cache.

FAQ

How is ElastiCache Serverless priced?

Per GB-hour of data stored in the cache and per ElastiCache Processing Unit (ECPU) consumed by reads and writes. It scales automatically with no nodes to provision, so you pay for the data held and the compute your operations use.

What is an ECPU?

An ElastiCache Processing Unit, the compute-billing unit for ElastiCache Serverless. Reads and writes consume ECPUs based on how many operations you run and their size, so the compute line scales with cache activity.

Is ElastiCache Serverless cheaper than nodes?

For spiky, unpredictable, or often-low traffic, usually yes, because you avoid paying for a large provisioned node sitting idle. For a large cache under steady, heavy load, a right-sized provisioned node, especially reserved, can be cheaper per unit.

How do I reduce ElastiCache Serverless cost?

Keep the working set lean so data-stored charges stay low, use sensible TTLs so the cache does not retain unused data, make operations efficient to reduce ECPU consumption, and ensure the cache deflects more downstream cost than it consumes.

When should I use provisioned nodes instead?

For large caches under steady, predictable, heavy load, where a right-sized provisioned node with a reservation is cheaper per unit than the Serverless data-stored and ECPU dimensions. Serverless is best when traffic is variable or the cache is often idle.

Does C3X estimate ElastiCache cost?

C3X prices provisioned ElastiCache clusters from your Terraform directly. For Serverless, cost depends on data stored and ECPUs consumed, which are usage inputs you model, since they scale with traffic rather than a fixed node size.

What to do next

Compare Serverless and node-based caching before you choose. C3X reads your Terraform and prices ElastiCache and the rest of your stack against a live catalog. Start with the quickstart.