CloudZero now splits an Amazon Bedrock line item into spend by model and by the team, user, or project behind each call, straight from your AWS bill.
Why this matters
When multiple teams, apps, and users all run inference through shared Amazon Bedrock endpoints, your AWS bill tells you what you spent on Bedrock, but not on what, or for whom. That’s the question you get asked the moment AI spend starts climbing.
Now you can answer it: which team is consuming the most on AI models, how your customer-facing chatbot compares to your internal summarization tool, and whether a given spike charges back to the right department or project. By tying each call to the IAM identity behind it, CloudZero turns “AI is getting expensive” into a number with a name on it — so chargeback and showback become real, and the usage patterns worth optimizing actually surface instead of hiding in a shared bucket.
What we built
AWS records the caller identity and project for every Bedrock inference call in the Cost and Usage Report (CUR 2.0). CloudZero now reads those fields and surfaces them in Explorer as dimensions you can group and filter on like any other tag, alongside the per-model usage Bedrock already reports. The result: break Bedrock spend down by model, by the IAM principal (the role or user that invoked the model), or by project, and combine those with the dimensions you already track.
How it works
This draws on CUR 2.0 (AWS Data Exports), so the prerequisite is that your account delivers a CUR 2.0 report into CloudZero. Once it’s flowing, CloudZero attributes each Bedrock line item to its caller identity and project automatically – no extra instrumentation, agents, or code changes, because the data comes from the AWS bill itself. The breakdown then appears as standard Explorer dimensions you can group, filter, and save into Views. Project attribution depends on a project cost-allocation tag being set on the workload; model and caller-identity breakdowns work as soon as Bedrock activity shows up in the report.
Read the docs to set up your CUR 2.0.
