Google has introduced new billing options and cost controls for AI agent workloads in Gemini Enterprise, with the changes also extending to developer tools including Google Antigravity and Android Studio.
The update adds a pay-as-you-go option alongside existing per-user subscriptions, as well as spend caps, anomaly alerts and a savings plan tied to monthly usage. The measures are aimed at companies trying to manage AI project costs as agent-based systems take on more work across business and engineering teams.
Under the new pricing structure, customers can keep fixed monthly seat subscriptions for workers with regular AI usage or use a consumption-based model for teams whose demand rises and falls. The pay-as-you-go edition of the Gemini Enterprise app is available to select customers and is due to roll out more widely.
The option removes upfront commitments and charges based on compute and token use at standard model API rates. It is intended to reduce the risk of companies paying for unused licences when project demand falls.
Another change affects how quotas are managed across tools. Usage from Google Antigravity, the Gemini Enterprise platform and the Gemini Enterprise app will now be combined into a single project-wide pool rather than handled through separate licences and billing arrangements.
For customers using Gemini Enterprise subscriptions, access to Google Antigravity and Android Studio AI features is being included under the same commercial framework for eligible users. This will allow development teams to draw from capacity already assigned to a Google Cloud project instead of managing separate product entitlements.
Spend controls
Google is also adding financial guardrails in the Google Cloud Billing Console. Companies can set hard monthly spending limits for AI projects, receive automatic email alerts when they reach 50%, 80% and 100% of budget, and pause agent API calls when a cap is reached.
If a project exceeds its cap, administrators can either resume activity manually or allow overages to continue under pay-as-you-go pricing. Overage spending can draw against a Flexible Savings Plan where one is in place.
Google is also introducing early anomaly detection for AI spending. The feature flags projects whose spending rises above normal levels and identifies the top three stock-keeping units behind the increase, giving finance and engineering teams a more detailed view of what has changed.
Alongside those controls, Google is promoting its pricing calculator as a way for organisations to estimate expected costs before launching projects. Customers can use it to model spending across user licences, developer tools and background agent runtimes.
Savings plans
A central part of the update is the Gemini Enterprise Flexible Savings Plan, which offers lower token prices in return for a monthly spending commitment. Customers that commit for one year receive a 10% discount, while those choosing a three-year term receive a 20% discount.
Google said there are no minimum or maximum spend requirements, allowing companies to set commitments that match current traffic and adjust them over time as AI usage grows. The plan is already available to self-serve customers and those on enterprise agreements.
The savings plan is designed to work across Gemini Enterprise usage and draw down against an existing Google Cloud enterprise agreement. That structure may appeal to larger companies seeking dedicated AI budgets without creating a separate procurement track.
Google also outlined a deferred-execution pricing model for certain workloads, though the feature is not yet broadly available. Eligible jobs can be marked for later processing and run during off-peak capacity windows, which could reduce inference costs by as much as half.
The deferred model is meant for work that does not need immediate completion. Such jobs would also bypass standard quota limits, potentially allowing businesses to run more agent-based activity within the same budget envelope.
FinOps focus
The announcement reflects a wider shift in enterprise AI buying, as the debate moves from access to models towards the economics of sustained use. As more companies test agents for software development, operations and internal productivity, finance leaders are looking for clearer visibility into token consumption, project overruns and the split between fixed and variable AI costs.
Google framed the changes around that concern, particularly for organisations trying to balance experimentation with tighter cost discipline. It is increasingly positioning Gemini Enterprise not just around model access, but also around budget controls and reporting features that procurement, finance and platform teams can use alongside developers.
In addition to reporting tools in the billing console, customers can use centralised billing reports with a FinOps agent to generate natural-language summaries explaining where AI budgets were spent. The aim is to help teams present returns on AI spending to senior management.
The new package also signals Google's attempt to integrate AI development tools more tightly into its wider cloud commercial model. By combining quotas, subscriptions and consumption charging across end-user applications and technical tools, it is seeking to make Gemini Enterprise a broader purchasing vehicle for both business users and software teams.
For customers already grappling with unpredictable AI demand, the most immediate effect may be the introduction of hard project caps and a second pricing route beyond seat-based subscriptions. Those mechanisms give companies a way to limit exposure when usage spikes while still allowing teams to continue work if administrators choose to permit overages.
Projects that hit their monthly limit will have agent API calls temporarily paused, protecting the budget without affecting the rest of a customer's production infrastructure.