# Billing on Confluent Cloud for Apache Flink Confluent Cloud for Apache Flink® is a serverless stream-processing platform with usage-based pricing. Confluent charges you only for the duration that your queries run. You configure Flink by creating a Flink [compute pool](compute-pools.md#flink-sql-compute-pools). Confluent charges you for the [CFUs](#flink-sql-cfus) you consume when you run statements within a compute pool. The compute pool itself scales elastically to provide the necessary resources; your cost depends on the actual CFUs you use per minute, not on the provisioned size of the pool. You can configure the maximum size of a compute pool to limit your spending. ## CFUs A CFU is a logical unit of processing power that Confluent Cloud for Apache Flink uses to measure the resources it consumes. Each Flink statement consumes a minimum of 1 CFU-minute but can consume more depending on the needs of the workload. #### NOTE If you’re a Flink Advantage Package customer, you can run up to four statements per CFU. ## CFU billing Confluent bills you for the total number of CFUs you consume inside a compute pool per minute. Usage appears in hours to apply hourly pricing to minute-by-minute use. For example, 30 CFU-minutes is 0.5 CFU-hours. CFU pricing : $0.21/CFU-hour, calculated by the minute ($0.0035/CFU-minute) Prices vary by [cloud region](../../billing/billing-dimensions.md#cloud-billing-regional-multiplier). ## Networking fees Using Flink to read and write data from Apache Kafka® doesn’t add any new Flink-specific networking fees, but you’re still responsible for the Confluent Cloud networking rates for data that you read from and write to your Kafka clusters. These are existing Kafka costs, not new charges that Flink creates. ## Cost management You can’t define the number of CFUs an individual statement requires. Confluent Cloud for Apache Flink counts CFUs for you. You can configure the maximum size of a compute pool to limit your spending by setting a parameter named MAX_CFU, which sets an upper limit on the hourly spend on the compute pool. If the size of the workload in a pool exceeds MAX_CFU, Confluent Cloud for Apache Flink rejects new statements. Existing workloads continue running but can experience increased latency. #### NOTE You can increase the MAX_CFU value after you create a compute pool, but you can’t decrease the initial MAX_CFU value. For more information, see [Update a compute pool](../operate-and-deploy/create-compute-pool.md#flink-sql-manage-compute-pool-update). You can set a minimum CFU floor for an individual statement with a [Baseline CFU](autopilot.md#flink-sql-baseline-cfu). A statement with a Baseline CFU is billed for at least that many CFUs even when it’s idle. Usage above the floor is billed on top, so setting a Baseline CFU can increase your CFU usage. For more information, see [Set a Baseline CFU for a Flink SQL Statement](../how-to-guides/manage-baseline-cfu.md#flink-sql-manage-baseline-cfu). For more information on CFU prices, see [Confluent Cloud Pricing](https://www.confluent.io/confluent-cloud/pricing/). ## Pricing examples Data streaming is a real-time business, and data streams oscillate on a minute-by-minute basis, creating peaks and troughs of utilization. You don’t want to allocate and overpay for processing capacity that you aren’t using. With Confluent Cloud for Apache Flink, you pay only for the processing power that you actually use. The following examples provide additional detail on how pricing works when processing streams using Confluent Cloud for Apache Flink. ### Data exploration and discovery Most SQL queries are short-running, interactive queries that help software and data engineers understand the streams they have access to. Querying the streams directly is an important and necessary step in the iterative development of applications and pipelines. In the following example, one user executes five different queries. Unlike other Flink offerings, Confluent Cloud for Apache Flink’s serverless architecture charges you only for the five minutes when these queries run, with all users able to share the resources of a single compute pool. It doesn’t matter whether the same person executes these queries, whether five different people execute them at the same time, or whether (as shown in the following diagram) the queries run at different points in the hour.  Example pricing calculation : - Number of queries executed = 5 - Total CFU-minutes consumed = 5 - Total charge: 5 CFU-minutes x $0.0035/CFU-minute = **$0.0175**
**Note:** The charge appears on the invoice as “0.083 CFU-hours x $0.21/CFU-hour”. ### Many data streaming applications and statements A data streaming architecture comprises many applications, each with its own workload requirements. An architecture can be a mix of interactive, terminating statements and continuous, streaming statements. Confluent Cloud for Apache Flink automatically scales the processing power of the Flink compute pool up and down in real time so that your applications have the processing power they need, while charging only for the minutes you use. In the following example, five streaming statements run in a single compute pool. The data streams oscillate, and you can see spikes of utilization for short periods within the hour. Each statement attracts a minimum price of 1 CFU-minute ($0.0035 in this example), and Confluent Cloud for Apache Flink automatically scales the statement up and down on a per-minute basis as needed.  | Statement | CFU-minutes | Statement Type | |-------------|---------------|------------------| | Q1 | 5 | Interactive | | Q2 | 60 | Streaming | | Q3 | 110 | Streaming | | Q4 | 10 | Interactive | | Q5 | 124 | Streaming | | Total | 309 | | Example pricing calculation : - Number of statements executing = 5 - Total CFU-minutes consumed = 309 - Total charge: 309 CFU-minutes x $0.0035/CFU-minute = **$1.0815**
**Note:** The charge appears on the invoice as “5.15 CFU-hours x $0.21/CFU-hour”. ## Related content - [Compute Pools](compute-pools.md#flink-sql-compute-pools) - [Confluent Cloud Pricing](https://www.confluent.io/confluent-cloud/pricing/)