Google's new spend caps stop runaway AI bills, but not much else.

Google Cloud has moved its 'spend caps' feature from preview to general availability. This lets customers set a monthly spending limit on specific services within a project. Once that limit is hit, the service blocks new requests, but leaves existing resources running and doesn't delete anything. The release adds an API for automation, faster enforcement for AI services based on live cost estimates, and automatic resets at the start of each month.

In practice, this is a very narrow tool designed for one specific nightmare scenario: a runaway AI model racking up enormous bills. It only works for Gemini API, Cloud Run, and a couple of other services—Compute Engine, storage, and networking are excluded. It covers a single project and service per month, ignoring folders, organizations, and multi-service budgets. The cynical takeaway is classic Google Cloud: a well-marketed feature that addresses a high-profile pain point (AI sticker shock) while leaving the vast, complex landscape of cloud cost management fundamentally unchanged.