Datadog pricing explained: what you actually pay

All prices below are Datadog's published list prices in US dollars, taken from datadoghq.com/pricing in September 2026, billed annually unless stated. Datadog changes its pricing from time to time, and your contract may carry different rates, so check the live page before you build a budget on any of these numbers.

How Datadog bills

Datadog is not one product with one price. Each product has its own meter: hosts for infrastructure, gigabytes and events for logs, unique tag combinations for custom metrics, hosts and spans for APM, sessions for RUM, test runs for Synthetics. You pay for the products you switch on, at the volume you send.

Hosts

Infrastructure Monitoring and APM are priced per host. Datadog does not bill on the host count at the end of the month. It takes the hourly host count across the month, discards the top 1 per cent of hours, and bills the highest remaining hour. A short scale-out event is forgiven. An autoscaling group that sits high for a few days is not.

Containers are allotted per host: 5 on Pro, 10 on Enterprise. Dense Kubernetes nodes can run past that, and the containers beyond the allotment are billed separately.

Logs: ingested and indexed are two different bills

This is the part of Datadog pricing most teams underestimate. Every log sent to Datadog is ingested, at $0.10 per GB. Only the logs you index become searchable in the Log Explorer, and indexing is billed per million events, at $1.70 per million for 15-day retention.

Indexing is where the money goes. A gigabyte of typical application logs is somewhere around a million events, so a gigabyte that costs $0.10 to ingest costs roughly $1.70 more to index for 15 days. Longer retention costs more per event. Logs that are ingested but not indexed can still feed metrics, archives and Flex Storage, which is priced per million events stored rather than per million indexed.

For the full breakdown of log billing, including retention and archive options, see our guide to Datadog logs pricing.

Custom metrics: billed by combination, not by name

Datadog defines a custom metric as a unique combination of a metric name and its tag values, including the host tag. One metric called checkout.latency tagged with 20 services and 50 hosts is not one custom metric. It can be up to a thousand.

Pro includes 100 custom metrics for every host monitored, and Enterprise includes 200. Above that, the list price is $5 per 100 custom metrics per month, and the billable count is the average number of custom metric hours across the month. Add one high-cardinality tag, such as a user ID, a request ID or a pod name, and a metric that was costing nothing can quietly become one of the larger lines on the invoice.

We cover cardinality and how to control it in Datadog metrics cost optimisation.

APM: hosts, then spans

APM is billed per APM host, from $31 to $40 per month depending on the tier. Each APM host comes with 150 GB of ingested spans and 1 million indexed spans a month. Above those allotments, ingested spans are billed per GB and indexed spans per million, at $1.70 per million for 15-day retention. The host fee is the predictable part. The span volume depends on traffic and on the retention filters someone set, or did not set.

RUM and Synthetics

Real User Monitoring is billed per 1,000 user sessions, with the rate depending on which RUM capabilities are enabled, so the live pricing page is the reference. Synthetic tests are billed per run: $5 per 10,000 API test runs and $12 per 1,000 browser test runs. Frequency multiplies both. A browser test that runs every minute from five locations is over 200,000 runs a month.

Annual commitments and on-demand

Datadog sells most products on an annual commitment, month-to-month, or on-demand, and the annual commitment is the cheapest of the three. Infrastructure Pro is $15 per host per month billed annually and $18 on-demand. Enterprise is $23 against $27. Standard log indexing at 15-day retention is $1.70 per million events annually and $2.55 on-demand. Committing to too little leaves usage on the higher rates. Committing to too much pays for capacity nobody uses.

The three things that drive most Datadog bills

The host price is the number people quote. It is rarely the number that makes a bill large. In the Datadog bills we review, three lines do most of the work.

1. Log indexing volume

The most common indexing strategy we see is no strategy at all: everything ingested gets indexed, at the default retention, with no exclusion filters. Debug output, health checks, load balancer access logs and verbose Kubernetes system logs all end up searchable for 15 days, and almost nobody searches them.

2. Custom metric cardinality

Custom metrics grow with tags, not with metrics. A library emits metrics by default, a team adds a tag to debug one incident, and the tag stays. Cardinality is invisible until the invoice, because nothing in day-to-day use shows how many billable combinations a tag has created.

3. APM host count

APM is priced per host, so it scales with infrastructure rather than with the number of services anyone actually traces. At $31 a host on top of a $15 Infrastructure Pro host, tracing every host roughly triples the per-host line, including hosts in development and staging where nobody reads the traces.

The long version, cause by cause, is in Why is Datadog expensive?

A worked example, built from list prices

This is illustrative, not a quote. It uses only the published annual list prices above, and one assumption: an average log event of about 1 KB, so each GB is roughly a million events.

Take an environment of 20 hosts on Infrastructure Pro, with APM on all 20, sending 1,000 GB of logs a month and indexing all of it at 15-day retention, and emitting 7,000 custom metrics against an allotment of 2,000 (100 for each of the 20 hosts). Span volumes are assumed to sit inside the APM allotment.

The line people budget for, hosts, is $300. The line that dominates is log indexing, at $1,700.

Now index only the quarter of those logs that anyone actually searches, 250 million events, and keep the rest in Flex Storage at $0.05 per million. Indexing falls from $1,700 to $425, Flex Storage adds $37.50, and the monthly total drops from $2,970 to $1,732.50. Nothing was switched off. The same logs are still collected. The difference is a decision about which of them need to be searchable at speed.

How to keep Datadog under control

None of this needs a different platform. It needs someone to own the decisions that the defaults otherwise make.

Indexing is a decision, not a default. Exclusion filters on low-value sources, index-level retention that matches what the logs are used for, and Flex Storage or archives for logs that need keeping but not searching. Logs that only ever feed a count can become log-based metrics instead.

Tags are reviewed before they ship. A tag with thousands of values belongs in logs or traces, not on a metric. Datadog's Metrics without Limits lets you choose which tags stay queryable on a metric, which cuts the number of indexed custom metrics you pay for.

APM goes where it earns its place. Full tracing on the services that page someone at 3am. Lighter sampling, or none, on the environments and endpoints where nobody reads the traces.

Usage is watched like any other production signal. Datadog's usage metering and estimated usage metrics can be monitored, attributed by team and service, and alerted on, so a spike shows up in a week rather than on the invoice.

The commitment matches real usage. Put the usage page next to the contract before renewal, not after it.

When to bring us in

We keep Datadog running for teams that depend on it, and part of that is making it cost what it should. We never recommend cutting coverage your on-call engineers rely on.

If the bill needs fixing now, our Datadog cost optimisation work finds where spend is growing without a matching gain in signal, fixes it at the source, and puts governance in place so it stays fixed.

If the question is where to start, a HealthScan is an independent, read-only review of your Datadog environment in one to two weeks. It covers cost and usage alongside coverage, alert noise and tagging, and ends with a health scorecard and a prioritised improvement backlog.

Deeper reading

• Why is Datadog expensive? The common causes of cost growth, in detail.
• Datadog logs pricing: ingestion, indexing, retention and archives.
• Datadog metrics cost optimisation: cardinality and custom metric governance.
• Datadog cost management articles: the rest of our writing on Datadog cost.