CloudWatch Omni Pricing: Why Telemetry Costs Will Drive Adoption

What AWS launched

AWS has introduced CloudWatch Omni, a new observability experience that pulls agent, application, and infrastructure telemetry into a single application-centric view. The goal is to help teams understand why AI agents behave the way they do in production 1. AWS's argument is unusually candid. It says conventional monitoring, including its own CloudWatch service, struggles to explain agent decisions. Today, engineers have to jump between three kinds of tools to piece together what happened [1]:

• Agent-focused tooling such as Amazon Bedrock AgentCore
• Application performance monitoring
• Infrastructure metrics

Omni is positioned as the fix for that fragmentation. AWS describes it as a purpose-built system for observing, evaluating, and experimenting with AI agents across any model provider, framework, or runtime 3. It is built on open standards and uses an eval-driven workflow. Built-in evaluators score agent output for quality, correctness, and coherence 3.

The product is delivered "off-console." Developers reach it from inside their IDE or through a standalone web experience, separate from the AWS Management Console 3. Investigations can also be run through natural language queries 1.

The pricing model is the real story

The feature pitch is compelling, but the commercial structure deserves closer attention. Omni has no per-seat or per-user license. Charges are tied to three dimensions of data [2]:

• Ingest: telemetry sent into CloudWatch
• Store: data retained over time
• Analyze: queries run inside Omni

Ingestion is billed per GB on arrival at standard CloudWatch rates [2]:

• Application and custom logs: $0.50/GB
• OpenTelemetry metrics: $0.50/GB, or $0.08/GB when they come from EKS Container Insights
• AWS service (vended) logs: tiered from $0.50 down to $0.05/GB
• Traces and spans: tiered from $0.35 down to $0.15/GB

Storage is charged per GB-month. Optional Intelligent Tiering spreads data across three storage tiers 2. Analysis is billed per GB scanned for logs and traces and per million samples for PromQL queries. Usage is only charged above an included allowance, and dashboards and alerts are bundled in 2. Existing CloudWatch Free Tier allowances continue to apply 2.

Why telemetry, not seats, will decide adoption

Removing seat licensing helps adoption. A platform team can roll Omni out to every developer's IDE without negotiating headcount-based contracts. That fits AWS's broader pitch of meeting engineers where they already work 3.

The catch is that agentic workloads are unusually chatty. A single agent run can produce several kinds of data:

• Model calls and tool invocations
• Retrieval steps and retries
• Evaluation records

Each of these may generate spans and log lines. When the bill scales with gigabytes ingested, retained, and scanned, Omni's cost tracks telemetry volume rather than team size. That volume can grow quickly as more agents move into production. Analysts quoted by InfoWorld flag rising telemetry costs as one of the main factors that could limit Omni's appeal 1.

The tiered rates for vended logs and spans suggest AWS expects heavy users and wants to reward scale 2. The included query allowance and bundled dashboards also lower the barrier for light usage 2. Even so, the economics push customers toward deliberate choices about sampling, retention windows, and query discipline. Teams that send everything at full fidelity and keep it indefinitely will probably see bills climb faster than those that design their telemetry pipelines with cost in mind.

Lock-in versus open standards

The other concern analysts raise is lock-in 1. AWS's messaging emphasizes open standards and support for any framework or model provider 3, and OpenTelemetry metrics are billed as a first-class ingestion type 2. In principle, that makes instrumentation portable.

In practice, the value Omni adds sits on top of CloudWatch storage and query billing. That includes the unified application-centric view, the evaluators, and the natural language investigation tools. Data may be collected in an open format, but the experience that makes it useful remains AWS-specific. Enterprises running multi-cloud or hybrid agent deployments will need to weigh that trade-off.

The bottom line

Omni is a credible attempt to make agent behavior explainable without forcing engineers across a patchwork of consoles. Its strongest selling points are the unified context and the eval-driven workflow 13.

The likely deciding factor for most buyers is less about features and more about how well they can forecast and control telemetry volume. Organizations that already know their CloudWatch spend and have mature data-retention practices are best placed to benefit. Others should model ingestion and query costs against realistic agent traffic before committing, because Omni's bill grows with the amount of data teams produce 2.