You added CloudWatch. You enabled X-Ray. You deployed to production. Then the bill arrived.

Observability on AWS is deceptively easy to start and surprisingly expensive to run. Most developers instrument their applications without understanding the pricing model underneath — and discover the hard way that logs, metrics, traces, and transaction search each have their own cost dimensions that multiply at scale.

In this talk, we'll build a clear mental model from the ground up. First, we'll cover why observability matters for AWS developers — where managed services make production systems opaque and why you need telemetry to operate them safely. Then we'll trace exactly how telemetry data flows from your code to a backend, and where OpenTelemetry fits into the AWS ecosystem today. Finally, we'll walk through each cost dimension — CloudWatch Logs ingestion, metric cardinality, X-Ray sampling, Transaction Search, and data transfer — with concrete pricing examples and the traps that surprise teams at scale.

You'll leave with a pricing mental model and practical OTel Collector configurations you can apply immediately to reduce your observability spend without losing signal.