Cloudflare said on October 1, 2026 that it is launching Cloudflare K2, a serverless event streaming service, in public beta on its Developer Platform. The company described K2 as a durable event streaming primitive that stores events sent to a stream as an ordered log for consumers to read at their own pace.
What K2 does
According to Cloudflare, producers send events to a K2 stream, which stores them as an ordered log. Consumers can read them in several ways, including splitting reads across a set of consumers or delivering all messages to all consumers.
The company said the service is fully serverless, scales to vast quantities of data and supports long-term retention, so that even long periods of consumer downtime do not lose data.
Cloudflare framed the product as an answer to a limitation of traditional Remote Procedure Call architectures, in which producers and consumers must align in scale and in time and events are dropped when consumers or downstream services become unavailable.
Built on R2 object storage
Under the hood, Cloudflare said K2 implements a partitioned, durable log on top of R2 object storage, which it said allows the service to scale to huge volumes of storage. The company said it originally built K2 as a durable buffer on the edge, initially to serve as the ingestion layer for Basin Pipelines.
Cloudflare said its edge spans a huge number of servers across over 335 cities, and that its architecture often cannot run traditional distributed systems software like Kafka. Relying on R2, which it said combines extremely durable storage (11 9s) with strongly consistent APIs, let it offload replication and consensus to the storage layer.
Because object stores do not support appends, Cloudflare said it accumulates writes in memory on an edge service and then writes events as a segment file, using R2's atomic operations to achieve ordering and strictly incrementing offsets without a separate coordination service. The company said the trade-off is higher produce latency: about 1 second at the 99th percentile in the initial release.
How it compares to Queues and Pipelines
Cloudflare said Queues are designed around tracking individual items of expensive or time-consuming work, with retries, delays and dead-letter queues, while K2 is designed for high-scale data movement, long-term retention and fan-out consumption. Messages in K2 are produced and consumed as batches, which the company said enables efficient processing at the expense of message-level retries.
For Basin Pipelines, Cloudflare said it recommends Pipelines when the end result is writing events to object storage or Iceberg tables, and K2 when doing custom processing or writing to other destinations.
Availability, limits and pricing
Cloudflare said K2 is available in public beta for accounts with Workers Paid subscriptions, with a maximum of 10GB of storage used and 30 MB/s produce per stream. Users who need higher limits are directed to the team on Discord or a limit increase form.
Usage will not be billed during the beta period, the company said. Once billing begins, Cloudflare said it anticipates charging $0.04 per GB for data produced, $0.04 per GB for data consumed and $0.02 per GB per month for data retained.
On the roadmap, Cloudflare listed higher write parallelism up to multi-GB/s streams, message keys and key-based ordering guarantees, push-based worker consumers, an express tier with lower produce and end-to-end latencies, and drop-in support for Apache Kafka clients. The company said it will share more details on K2's design in an upcoming technical deep dive.
What to do
- Create a stream first: Cloudflare says streams can be created via cf, Wrangler, the dashboard or the API, for example with the command cf k2 streams create –name app_events –http-enabled.
- Produce events to the stream using either the HTTP API or a Worker binding; Cloudflare notes K2 represents data as bytes, so any format or encoding can be used.
- Create a subscription through the HTTP API to divide work between consumers and enable read parallelism, then poll it from each consumer with the consume endpoint.
- After consuming a batch, ack it to mark it processed, nack it to request redelivery, or extend the lease if more processing time is needed within the five-minute lease window.
- For pub/sub-style delivery, Cloudflare says you can create a separate subscription for each consumer so every consumer sees all messages, or mix both approaches.
- Cloudflare directs users who need higher beta limits to contact the team on Discord or fill out the limit increase form, and points to the K2 docs for full API details.
Key facts and where they come from
- Cloudflare launched K2 in public beta as a durable event streaming primitive on its Developer Platform.
Today we are launching Cloudflare K2 in public beta to solve this problem. K2 is a durable event streaming primitive on the Developer Platform.
- K2 is implemented as a partitioned, durable log on R2 object storage.
Under the hood, K2 implements a partitioned, durable log on top of R2 object storage, which allows it to scale to huge volumes of storage.
- Initial release adds about 1 second of produce latency at the 99th percentile.
In our initial release of K2, this adds up to about 1 second of produce latency at the 99th percentile of response times.
- Beta limits are 10GB of storage and 30 MB/s produce per stream, for Workers Paid accounts.
K2 is available today in public beta for accounts with Workers Paid subscriptions, within these limits: Maximum of 10GB of storage used 30 MB/s produce per stream
- Anticipated pricing after the beta is $0.04/GB produced, $0.04/GB consumed and $0.02/GB/month retained.
Data Produced $0.04 / GB Data Consumed $0.04 / GB Data Retained $0.02 / GB / month
- Consumers receive a five-minute lease on a batch and can ack, nack or extend it.
When a client calls consume, they receive a lease for that particular batch of events for 5 minutes.
- Cloudflare's edge spans servers across more than 335 cities.
Pipelines runs on the Cloudflare edge, which spans a huge number of servers across over 335 cities