A Microsoft offering that enables tracking of cloud usage and expenditures for Azure and other cloud providers.
Hi Michał Duda & thx for join me at Q&A platform,
this can happen with Foundry, especially if u accidentally created a managed compute deployment while experimenting with models. Unlike pay-per-token models, managed compute can charge for allocated GPU capacity even when nobody is actually using the model. So not sending any requests doesn't necessarily mean there won't be charges.
First, I'd check Cost Management + Billing > Cost analysis and group the charges by Resource and Meter. That should help identify what was actually billed. Three entries on an invoice don't automatically mean u were charged three times for the same thing. They could represent different usage periods, resources, or billing meters.
Also check ur Foundry deployments and any Azure Machine Learning managed online endpoints. If something is still deployed, delete it to stop further compute charges. Just closing Foundry or leaving the project unused won't necessarily stop billing. For the refund, open an Azure Billing support request through Help + support. Billing support is free, even without a paid support plan. Explain that u were exploring Foundry, didn't realize a billable compute resource had been provisioned, and never intentionally used the service. Ask for a one-time courtesy refund or billing adjustment.
Include the invoice number, three charge amounts, resource IDs, and the dates involved. Microsoft can review the actual usage and decide whether a refund is possible, but it's not guaranteed.
Btw, if u share the exact service and meter names shown on the invoice (without personal billing details), we can probably figure out what generated those charges before u contact support.
rgds,
Alex
p.s. if my answer is hepl - pls mark it as an answer