Deepgram adds billing and GPU metrics for SageMaker speech deployments

Deepgram's SageMaker AI deployments now publish billing, usage and engine telemetry into customers' CloudWatch accounts, AWS said. The Enhanced Metrics feed reports the consumed units used for AWS Marketplace metered billing and breaks usage down by category, model, transport and enabled features. AWS also said Prometheus and OpenTelemetry support provides engine, per-GPU and host metrics through SageMaker detailed observability. The metrics use the existing logging path, so the container does not need outbound network access or additional permissions.