Build a clearer cost strategy for production AI agents

Microsoft outlined an AI FinOps approach for Microsoft Foundry focused on moving AI projects from pilots to measurable returns.

The approach combines cost attribution, model routing, caching, fine-tuning, workflow optimization, token limits, quotas, and budget governance. These controls are intended to improve visibility into production agent costs and reduce unnecessary spending.

Right-size licensed and CPU-bound workloads with Azure VM vCore customization

Azure VM vCore Customization is now generally available, with options to disable simultaneous multithreading and configure constrained cores.

These controls can help teams right-size licensed or CPU-bound workloads by avoiding payment for unused virtual CPU capacity. The result is more control over the compute resources assigned to workloads where licensing or CPU requirements don’t match the default configuration.