Amazon Bedrock's New Granular Cost Attribution: Track Your AI Spending Like Never Before

As AI inference becomes a major part of cloud spending, organizations are struggling to understand who's using what and how much it's costing. The good news? AWS just solved this problem with a game-changing feature for Amazon Bedrock users.

What's New: Automatic Cost Tracking by User

Amazon Bedrock now automatically attributes inference costs to the specific IAM principal (user, role, or federated identity) that made each API call. This means no more guesswork about AI spending—you'll know exactly who's using Claude, which models they're choosing, and how much each interaction costs.

The best part? This works out of the box with zero setup required. Your existing workflows don't change, but now you get detailed cost insights flowing directly into your AWS billing reports.

Real-World Example: See Your AI Spending in Action

Here's what you'll see in your Cost and Usage Reports (CUR 2.0):

Now you can instantly see that Alice prefers the cost-effective Sonnet model, while Bob is using the premium Opus model and generating more expensive output tokens.

Level Up with Tags: Team and Project Cost Rollups

Want to aggregate costs by team, project, or cost center? Simply add tags to your IAM users or roles. These tags automatically flow through to your billing data, giving you powerful aggregation capabilities.

For example, tag your data science team members:

aws iam tag-user \
  --user-name alice \
  --tags Key=team,Value="DataScience" Key=project,Value="ChatBot"

This creates billing entries that look like:

• Individual tracking: See exactly what each person spent
• Team rollups: Total spending for the entire DataScience team
• Project analysis: Compare costs across different AI initiatives

Four Common Scenarios: Choose Your Setup

AWS identified four main patterns for how teams use Bedrock, each with different attribution capabilities:

1. Developer Teams with Direct Access

Best for: Small teams, development environments
Setup: Each developer gets IAM user credentials or API keys
Result: Perfect per-user cost tracking with zero configuration

2. Applications with IAM Roles

Best for: Production applications, Lambda functions
Setup: Each app uses a dedicated IAM role
Result: Track costs by application or service

3. Enterprise Identity Integration

Best for: Large organizations using Okta, Entra ID, etc.
Setup: Federated identities with session tags
Result: User-level tracking without managing AWS credentials

4. LLM Gateway Pattern

Best for: Organizations with existing AI gateways
Setup: Requires per-user session management
Result: User attribution through your existing proxy layer

Getting Started: Your Next Steps

The beauty of this feature is its simplicity. Here's how to get started:

1. Enable IAM principal data in your CUR 2.0 export configuration
2. Add tags to IAM users/roles if you want team or project rollups
3. Activate cost allocation tags in the AWS Billing console
4. Wait 24-48 hours for tags to appear in Cost Explorer and reports

Why This Matters for AI Teams

This isn't just about billing—it's about enabling smarter AI operations:

• Cost Optimization: Identify heavy users and expensive model choices
• Chargeback Accuracy: Bill departments based on actual usage
• Financial Planning: Understand AI spending patterns for better budgeting
• Usage Analytics: See which models and features drive the most value

As AI becomes central to more business processes, having granular cost visibility isn't just nice to have—it's essential for sustainable AI adoption. Amazon Bedrock's new cost attribution feature makes this level of insight accessible to every organization, regardless of size or technical complexity.

Source: AWS Machine Learning Blog by Ba'Carri Johnson