AWS bill line item
USE1-Host:ml.g5.xlarge
Billed for every hour a SageMaker real-time inference endpoint's instance is running, regardless of whether it's actively serving prediction traffic.
SageMaker real-time endpoint hours
Verified against official AWS documentation last checked
Why it shows up
GPU endpoints are among the most expensive per-hour instance types available, and they're commonly left running around the clock after a demo, proof of concept, or project winds down, because nothing prompts anyone to delete the endpoint.
Endpoints are also frequently over-provisioned, sized or replicated across multiple instances for peak load that only happens occasionally, so most hours are billed at the full GPU rate for near-idle capacity.
How to cut it
- Delete or stop endpoints as soon as a project or experiment is done.
- Use serverless or asynchronous inference for spiky or low-traffic workloads instead of an always-on real-time endpoint.
- Enable endpoint auto-scaling, including scale-to-zero where supported, so idle periods don't bill at full capacity.
- Right-size the instance type/family to the model's actual latency and throughput needs.
How to read this code
USE1 / USE2us-east-1 (N. Virginia) / us-east-2 (Ohio)
:instanceType suffixThe specific instance type or size a charge applies to, e.g. BoxUsage:m5.large. One line per distinct type in use.
Source
Source: AWS docs →Found USE1-Host:ml.g5.xlarge on your own bill?
Drop your Cost Explorer CSV into the AWS Bill Analyzer. It'll flag line items like this one on your own bill and link straight back to the fix.
Analyze your own bill →Token spend moves every day.
Model prices and rate limits change between invoices. CostMon tracks your AI spend daily, so a change shows up before the bill does.