Event streaming
Each provider is profiled from its own pricing page, with the source recorded against every figure. Each has a standing monthly floor — $100, $385, about $547, and one that is not published — and on top of it one bills five meters at once, one bills three, and one bills a single number. The spread is architectural: a service that writes straight to object storage has no inter-zone replication to charge for, and its price list shows it.
Event streaming
| Provider | Entry paid plan | Partition limit | What is metered |
|---|---|---|---|
| Amazon MSK | $0.75 per cluster-hour on Serverless — about $547 a month for a cluster that has moved no data at all, before partitions, transfer or storage are counted. | 2,400 partition leaders per Serverless cluster for non-compacted topics, and 120 for compacted ones. Partitions are metered as well as capped — $0.0015 per partition-hour, so 2,400 of them cost about $2.65 a day carrying no traffic. | Serverless bills FIVE dimensions simultaneously: $0.75 per cluster-hour, $0.0015 per partition-hour, $0.10 per GB in, $0.05 per GB out and $0.10 per GB-month stored. Provisioned clusters swap the first two for a per-broker hourly rate. Nothing here is a single number. |
| Confluent Cloud | Basic gives the first eCKU free and charges $0.14 per eCKU-hour after it, so a small cluster starts near zero. Standard, the first tier with a 99.99% SLA, starts at about $385 a month. | No figure is published. Enterprise advertises "enhanced partition limits" over Standard without saying what either limit is, so the ceiling cannot be compared with a vendor that states one. | Three meters running together. Compute is billed per eCKU-hour — an elastic Confluent Kafka Unit that autoscales — data in and out is billed per GB, and data at rest per GB-month. The rates are published pre-replication, so a replicated topic costs a multiple of the figure shown. |
| Redpanda | Not documented | 5,000 partitions on Serverless, 112,500 on Bring Your Own Cloud. | Not published. The pricing page names two tiers, Serverless and Bring Your Own Cloud, and states throughput, partition and SLA ceilings for each — but no rate, no unit and no monthly figure appears anywhere on it. Serverless is described as pay-as-you-go or annual commit; BYOC as annual commitment only. |
| WarpStream | $100 a month for the Dev tier, which carries NO uptime SLA, before any per-gibibyte write charge. The first tier with an SLA is Fundamentals at $500. | 4,096 on Dev, 16,384 on Fundamentals, 50,000 on Pro and 200,000+ on Enterprise. Partitions are capped by tier and not billed. | Two lines, not one. A cluster-tier fee — $100 a month for Dev, $500 for Fundamentals, $1,500 for Pro — plus write throughput metered per gibibyte: $0.01/GiB to 50 TiB, falling to $0.003 beyond 6,250 TiB. Reads, partitions and storage are not billed; data sits in your own object storage. |
Which to consider
You are already on AWS and want Kafka inside the same IAM, the same VPC and the same bill. Serverless scales partitions for you and states no maximum retention at all, which nothing else here offers.
The reference implementation, and the one that shows its working. Three meters are each published with a rate, and the page states they are pre-replication — a disclosure the others skip entirely. Standard at about $385 a month is the cheapest 99.99% SLA in this category.
Knowing the ceilings before you talk to anyone. It publishes throughput, partitions and SLA for both tiers — 100 MB/s and 5,000 partitions on Serverless, 2 GB/s and 112,500 on Bring Your Own Cloud — which is more capability detail than the others put on a pricing page at all.
Write-heavy workloads where inter-zone replication is the bill. Data is acknowledged straight into your own object storage, so partitions, reads and storage are not metered at all and only gigabytes written are — $0.01 falling to $0.003 at volume.