Artie now runs Apache Kafka for you on AWS.
You pay AWS’s on-demand prices for the cluster, the same as running Kafka on EC2 yourself, plus 25% of the broker instance cost. For 3 r8g.large brokers with 1 MB/s in and 500 GB retained, that comes to $436 a month. The same cluster costs $1,110 on Amazon MSK.
Artie Managed Kafka replaces failed brokers, rebalances partitions, grows disks and rolls out upgrades one broker at a time. And our engineers are the ones on call.
Artie’s CDC pipelines run on Kafka, and we needed a Kafka we could tune and run in our customers’ clouds. Confluent locked broker settings behind an enterprise tier. Redpanda needed its agent inside each of our data plane clusters. So we initially chose Amazon MSK.
On MSK, we kept writing our own tools to work around its limits. Its disks could only grow, and config changes took over an hour. When we added brokers, they sat empty until our rebalancer moved partitions onto them.
The outage that settled our decision started at 9:24 pm on May 9, 2024. MSK’s IAM authentication spiked a broker’s CPU and trapped our consumers in rebalance loops. We added reconnect jitter that night and later switched to SASL/SCRAM. Over the next year, we moved every workload to our own Kafka on Kubernetes. We set the broker configs, and software handles rebalancing, disk growth and broker recovery.
Your bill has two parts:
Cluster Workload Artie Amazon MSK Savings Small: 3 × r8g.large 1 MB/s in, 500 GB retained $436/mo $1,110/mo 61% Medium: 6 × r8g.large 5 MB/s in, 2 TB retained $1,119/mo $2,740/mo 59% Large: 12 × r8g.xlarge 20 MB/s in, 8 TB retained $3,648/mo $10,950/mo 67%
us-east-1. AWS prices as of November 3, 2026; MSK prices from AWS’s October 14, 2026 price list. Assumes each message is read once, in the same region. Artie runs in one availability zone at launch (multi-AZ is on the roadmap). Amazon MSK runs the same number of brokers with the same memory across three zones, with its cross-zone client traffic.
Book a migration call. An Artie engineer plans the move with you, mirrors your topics with MirrorMaker, and then cuts over your producers and consumers. If you ever leave, you mirror your topics out the same way.
If you run Kafka today, price your cluster. Then create a cluster for dev or staging, point real traffic at it, and try to break it.