The new line on every cloud bill
Model APIs and managed ML platforms have gone from a rounding error to one of the fastest-growing lines on many cloud bills. They are also easy to lose inside a long service list. CloudMonitor separates AI services out of the cost data it already syncs, so you can see this month’s AI spend at a glance.
What is tracked
Amazon Bedrock and Amazon SageMaker charges from AWS Cost Explorer, and Azure OpenAI and Cognitive Services charges from Azure Cost Management. Because they are ordinary cost lines, they automatically count toward your budgets and month-end forecast, appear in the cost-by-service breakdown and are covered by daily anomaly detection — the check most likely to catch an unbounded batch job or a prompt loop early.
What it doesn’t do (yet)
CloudMonitor reports AI spend per service from your bill. It doesn’t ingest model invocation logs or token counts, so it can’t attribute spend to individual product features or teams. Google Vertex AI line items are recognised, but Google Cloud cost data requires a BigQuery billing export that CloudMonitor doesn’t read yet.
Keeping AI costs in check
The fundamentals are the same as for any other service: set a budget, watch the daily curve, and review the biggest movers every week. Our guide to FinOps for engineers explains how to build that habit without a dedicated team.