AWS Kinesis Data Firehose cost: per GB ingested, plus conversions

Quick answer

Kinesis Data Firehose bills about $0.029 per GB ingested for the first 500 TB per month, dropping in higher tiers, plus extra for format conversion (to Parquet or ORC) and for delivery into a VPC. Cost tracks ingested volume, so compressing and batching upstream and converting only when needed keep it down.

Firehose is the managed way to load streaming data into destinations like S3, Redshift, and OpenSearch. Its pricing is refreshingly simple: you pay per gigabyte of data ingested, with a couple of add-ons for optional processing. Volume is the lever.

Ingestion is the main line

The base charge is per GB ingested, around $0.029 for the first 500 TB in a month and lower in higher tiers. There is no hourly or shard charge as with Kinesis Data Streams, so an idle stream costs nothing and a busy one costs in proportion to the bytes it carries.

Optional add-ons

Dimension	Rate (approximate)
Ingestion, first 500 TB/month	~$0.029 / GB
Format conversion to Parquet/ORC	Additional per GB
VPC delivery	Per hour per AZ + per GB
Dynamic partitioning	Additional per GB + per object

Format conversion and dynamic partitioning are useful for making the landed data cheap to query in Athena, but they add to the Firehose bill, so enable them where the downstream query saving justifies the ingest cost.

Reduce ingested volume

Because cost is per GB ingested, shrinking the data before it reaches Firehose is the direct lever: compress records, drop fields you do not need downstream, and batch efficiently so overhead per record is low. Converting to Parquet in Firehose then makes the stored data cheap to query, trading a small ingest add-on for a large Athena scan saving.

Firehose vs Data Streams

Firehose has no shard management and bills purely per GB, which suits load-and-forget delivery. If you need low-latency, multiple consumers, or replay, Kinesis Data Streams (with its shard-hour pricing) fits better, the tradeoff covered in MSK versus Kinesis.

FAQ

How is Kinesis Data Firehose priced?

Primarily per gigabyte ingested, about $0.029 per GB for the first 500 TB per month and lower in higher tiers, plus optional add-ons for format conversion, dynamic partitioning, and VPC delivery. There is no shard or hourly charge.

How do I reduce Firehose cost?

Reduce ingested volume: compress records, drop fields you do not need downstream, and batch efficiently. Enable format conversion and dynamic partitioning only where the downstream query saving justifies the per-GB add-on.

Does Firehose charge for idle streams?

No. Firehose bills per gigabyte ingested with no shard or hourly fee, so an idle delivery stream costs nothing and a busy one costs in proportion to the data it carries. This differs from Kinesis Data Streams, which charges per shard-hour.

Is format conversion worth the extra Firehose cost?

Often yes. Converting to Parquet or ORC adds a per-GB Firehose charge but makes the landed data far cheaper to query in Athena, which bills per terabyte scanned. Enable it when the query saving exceeds the ingest add-on.

Firehose or Kinesis Data Streams?

Firehose suits load-and-forget delivery to destinations like S3 with no shard management, billing purely per GB. Data Streams suits low-latency, multiple-consumer, or replayable pipelines and bills per shard-hour. Match the model to your consumer needs.

Does C3X estimate Firehose cost?

Firehose cost is driven by ingested volume, a usage input. C3X prices the aws_kinesis_firehose_delivery_stream resource and its configuration, and you model ingest volume to estimate the per-GB charge.

What to do next

Price your streaming pipeline before you build it. C3X reads your Terraform and prices Firehose and the rest of your stack against a live catalog. Start with the quickstart.