AWS CloudWatch Completes OpenTelemetry Triad with Native OTLP Metrics and PromQL Support

Amazon Web Services has launched native OpenTelemetry Protocol (OTLP) metrics ingestion alongside PromQL query support in Amazon CloudWatch. This release completes CloudWatch's end-to-end integration across all three primary observability pillars—metrics, logs, and distributed traces—under vendor-neutral OpenTelemetry standards. In addition to native OTLP wire ingestion, the service introduces support for up to 150 metric labels (up from the legacy 30-dimension ceiling for custom metrics), bearer token authentication for environments without IAM credentials, cross-account metric centralization, and Amazon EKS Container Insights integration for hardware-level GPU and network monitoring. For platform engineers and SREs, this move significantly reduces the operational overhead of running parallel telemetry pipelines. Historically, cloud-native teams relying on Prometheus-style label querying and high-cardinality tags had to either deploy managed Prometheus clusters or run intermediate sidecars to reshape OTel metrics before shipping them to CloudWatch. The expansion to 150 labels directly addresses Kubernetes and microservice telemetry requirements, where container metadata, pod identifiers, and deployment revisions quickly breach restrictive dimensional limits. This update reflects the broader consolidation around OpenTelemetry as the de facto data-plane standard for cloud telemetry. Major cloud providers are shifting away from proprietary agent architectures in favor of open-wire protocols. As workloads become increasingly distributed across multi-cloud and edge environments, enterprises demand consistent collection frameworks that separate runtime instrumentation from visualization and storage backends. By embedding both OTLP ingestion and PromQL querying natively, CloudWatch aligns managed cloud monitoring with open-source engineering workflows. In practice, teams should evaluate streamlining their telemetry collection architecture by migrating custom metric exporters to direct OTLP streams. Organizations operating on Amazon EKS can deploy the unified CloudWatch Observability add-on to capture container metrics, distributed traces, and NVIDIA DCGM hardware statistics through a single collection layer. However, teams must institute cardinality governance and metric filtering at the OTel Collector level to prevent high-frequency, multi-label metrics from driving unexpected ingestion and storage costs at scale.