OpenTelemetry graduates CNCF as cloud giants adopt the standard
Something worth paying attention to happened in cloud infrastructure this week. OpenTelemetry, the open source framework for collecting traces, metrics, and logs from distributed systems, graduated from the Cloud Native Computing Foundation on May 21. The graduation itself is a maturity designation, but the more interesting story is what the three largest cloud providers are quietly doing around it.
Specifically, AWS, Microsoft, and Google are all adding native OpenTelemetry Protocol support into their monitoring services, and they are doing it at roughly the same time.
AWS brought OpenTelemetry metrics support to Amazon CloudWatch in public preview earlier this year, letting teams send metrics directly using OTLP without building custom conversion logic in between. On top of that, it added OpenTelemetry-based Container Insights for Amazon EKS, giving Kubernetes operators a standardized path for collecting cluster health data into CloudWatch.
Meanwhile, Microsoft documented OTLP ingestion for Azure Monitor across Kubernetes, virtual machines, scale sets, and Arc-enabled hybrid servers. Google, for its part, added OTLP support to Cloud Trace and Cloud Monitoring, and published a Google-built OpenTelemetry Collector for teams running on Cloud Run.
None of this, however, replaces the providers’ own monitoring tools. Azure Monitor, CloudWatch, and Google Cloud Observability all continue operating as they did before. What changes instead is the layer between applications and those backends. Because OpenTelemetry standardizes how telemetry leaves a workload, teams can instrument their applications once and route that data to whichever backend they need, without rewriting instrumentation every time something changes.
That distinction matters more as infrastructure grows more fragmented. After all, modern applications frequently span Kubernetes clusters, serverless functions, managed databases, and now AI agents running inside development environments. Getting consistent observability across all of that has historically required stitching together provider-specific agents and SDKs that do not always agree on format.
As a result, OpenTelemetry sits underneath all of that as a common collection and export layer. The fact that all three major providers are now building OTLP support into their native services suggests the standard has enough momentum that working against it no longer makes practical sense.
For platform engineers and cloud architects, therefore, the graduation is less about ceremony and more about what it confirms.

