Datadog vs CloudWatch 2026: when each wins

Datadog and CloudWatch describe two genuinely different cost philosophies. Datadog charges per host, with separate add-on meters for APM, log ingestion, log indexing, custom metrics, RUM, synthetics, and database monitoring. The per-host rate (around $18 list for infrastructure, $31 for APM) is the anchor, and each new host adds a known fixed cost. CloudWatch has no per-host concept at all. It is pure usage metering: you pay per custom metric, per gigabyte of logs ingested and stored, per dashboard above the free three, per canary run, per RUM event, and per gigabyte scanned by Logs Insights.

The Datadog model rewards homogeneous fleets and predictable observability. A fleet of similar hosts with similar APM coverage and log volume behaves predictably; you can forecast the bill from the host count. The model also bundles a great deal of capability into that per-host rate, so the marginal cost of turning on another dashboard or another integration is zero. The cost is the floor: even a lightly instrumented host carries the full per-host charge.

The CloudWatch model rewards restraint. There is no floor, so a tiny AWS workload can cost a few dollars a month. But there is also no ceiling and no bundle, so the bill is exactly the sum of every meter you enable. Custom metrics persist indefinitely once published. Logs ingestion at $0.50 per gigabyte is the dominant line item on log-heavy workloads. The model punishes teams that turn everything on without auditing what they query, and it rewards teams that keep their metric and log footprint disciplined.

The practical consequence: CloudWatch is almost always cheaper for AWS-only workloads, sometimes dramatically so at small scale, while Datadog narrows the gap at enterprise scale through negotiation and pays its premium back in capability. The decision is rarely about price alone.

The binary framing (CloudWatch or Datadog) is rarely how mature AWS teams operate. The data that is already in CloudWatch (AWS service metrics, VPC flow logs, Lambda logs, basic infrastructure telemetry) is cheapest to leave in CloudWatch, where ingestion is native and no egress or re-instrumentation is required. The data where Datadog earns its premium (application APM, distributed traces across microservices, polished dashboards for on-call teams, cross-cloud workloads) is worth routing to Datadog even at the higher per-host rate.

The instrumentation choice is what makes the hybrid clean. Applications instrumented with the AWS Distro for OpenTelemetry (ADOT) or vanilla OpenTelemetry can route traces to X-Ray and Datadog simultaneously, and can switch backends in a collector configuration change rather than an application rewrite. Teams that hard-wire the Datadog agent into every service lose that optionality and face a multi-month migration if the economics change. ADOT preserves the ability to keep CloudWatch as the cheap AWS-native layer while sending the high-value telemetry wherever it is most useful.

The cost discipline in a hybrid is to be deliberate about what goes where. Sending all CloudWatch logs into Datadog for indexing pays both the CloudWatch ingest meter and the Datadog indexing meter on the same data, the most common way a hybrid bill balloons. The pattern that works is to filter at source, keep high-volume low-value logs in CloudWatch, and forward only the subset that genuinely benefits from Datadog correlation.

All pricing in this comparison is verified against published vendor pricing pages in June 2026: datadoghq.com/pricing and aws.amazon.com/cloudwatch/pricing. CloudWatch prices vary by AWS region; the figures here are us-east-1 list. AWS X-Ray and CloudWatch Logs Insights are separate services on their own meters. Datadog discounts heavily on multi-year commitments above 500 hosts. Obtain current quotes before basing a decision on list pricing alone.