AutoMQ is an Apache-2.0 licensed, Java-based diskless distribution of Kafka that keeps stream data on S3 instead of broker-local disks, aimed at platform and infrastructure teams running event streaming at cloud scale.

What it is

AutoMQ is a rework of Apache Kafka in which the durability layer moves off broker-attached volumes and onto S3. The project describes itself as "Diskless Kafka® on S3" and sits in the Kafka and event-streaming ecosystem; the repository's topic list is automq, diskless, kafka, and serverless. The codebase is Java, released under the Apache-2.0 licence. The project's own description claims tenfold cost savings, no cross-AZ traffic cost, autoscaling in seconds, single-digit-millisecond latency, and multi-AZ availability.

The problem it addresses is structural. Conventional Kafka brokers pin data to local volumes, which forces capacity planning around peak load, keeps replicas spread across availability zones so replication traffic crosses zone boundaries, and turns scaling into a rebalancing operation instead of an instant adjustment. AutoMQ replaces that local-disk storage layer with object storage, so the durable copy of the log lives in S3 and a broker is no longer bound to the disk beneath it. The concrete artefacts replaced are the disk-backed Kafka broker and, for teams on managed services, the AWS MSK cluster that at least one published customer story describes migrating away from.

Key capabilities

• Diskless storage model places stream data on S3 rather than on broker-local disks, removing the broker as the holder of the durable log.
• Published comparison material covering AutoMQ versus Apache Kafka on both cost and performance.
• Claimed 10x cost savings and elimination of cross-AZ traffic cost.
• Autoscaling measured in seconds, positioned against the over-provisioning that fixed broker fleets require.
• Single-digit-millisecond latency with multi-AZ availability, per the project description.
• Apache-2.0 licence on a Java codebase, with 10,770 stars, 771 forks, and 60 open issues at the time of writing.
• Repository topics automq, diskless, kafka, and serverless define the project's scope.

Who uses it and how

• LG U+ runs 2.2 billion messages per day through AutoMQ to modernise log pipelines on AWS.
• JD.com replaces Kafka with AutoMQ at 40 GiB/s scale.
• Geely uses it for hybrid-cloud streaming across a 10-million-vehicle fleet.
• Honda reports cutting Kafka TCO by 50% with the cloud-native diskless architecture.
• Avia Games migrated off AWS MSK to remove maintenance disruptions, while Grab reports 3x data streaming efficiency and Tencent Cloud EMR ships AutoMQ as a first-party service.

Getting started

The README excerpt points readers to the project documentation and to an "Explore AutoMQ Playground" hosted option. It does not list a package name, Docker image, or compose file in the material available here, so installation detail must come from the documentation at automq.com.

How it compares

The facts name Apache Kafka and AWS MSK as the reference points, and AutoMQ publishes cost and performance comparisons against Apache Kafka while one customer story documents a migration away from AWS MSK. Apache Kafka is the open-source baseline it reworks; AWS MSK is the managed alternative some adopters are leaving. The distinction the project draws is a diskless, S3-backed deployment model against disk-backed brokers, whether self-run or provider-run.

When to use it — and when not to

A self-hoster still has to run brokers, own an S3 bucket, and manage access controls and scaling policy around it, so this is not an operational free lunch. Teams with no Kafka-shaped workload gain little, since the value sits in the Kafka storage layer specifically. Note also that the README excerpt is heavy on badges and customer links and light on installation instructions, that 60 issues remain open, and that the headline cost and latency figures are vendor-published claims rather than independently verified results in these facts.