Google Cloud has announced expanded billing flexibility and new cost-management controls for Gemini Enterprise and agent workloads. The 26 August update introduces a pay-as-you-go consumption edition for the Gemini Enterprise app, consolidated quota handling for developer tools, project-level spend guardrails and Flexible Savings Plans for qualifying usage. Together, the changes address a recurring enterprise concern: agent workloads can be valuable, but their variable use can make costs difficult to anticipate.
AI agents do not behave like a conventional per-seat software product. One team may use a predictable productivity assistant throughout a workday, while another runs a burst of model-intensive workflows only when a project reaches a particular stage. That difference creates a mismatch when every use case is priced in the same way. Google’s announcement is about giving organisations a choice between steady subscription-style access and consumption-based capacity, while keeping those options inside a shared administrative model.
New ways to pay for Gemini Enterprise use
The new Gemini Enterprise app pay-as-you-go consumption edition is available for select customers and is expected to roll out more broadly, according to Google Cloud. It has no upfront commitment or base subscription fee; customers pay for the compute and tokens their teams consume at standard model API rates. That may suit intermittent or experimental agent projects where allocating a fixed seat to each participant would leave capacity underused.
The established per-user seat subscription remains part of the mix. Google says it includes daily quota pools shared across a project, giving teams a predictable baseline for routine productivity work. The new structure means an organisation can combine that predictable access with usage-based capacity instead of choosing one model for every user and workload. The relevant trade-off is straightforward: fixed subscriptions can simplify budgeting for steady use, while consumption pricing can better match demand that rises and falls.
Google also says pooled quotas will be available for Antigravity in Gemini Enterprise. Business applications, developer tools and custom agents can draw from the same project-wide daily allowance. Pooled quota is used first, and administrators can decide whether overages are permitted, at which point they are charged at pay-as-you-go rates. For teams running several AI surfaces, that can reduce the chance that one allocation goes unused while another group has to buy separate capacity.
Controls intended to prevent cost surprises
The announcement puts equal weight on governance. Google says customers can set hard monthly caps on AI spend and projects, estimate agent runtime costs and identify sudden budget spikes before they appear on an invoice. These controls are meant to make experimentation less open-ended. A team can allocate a bounded budget, observe how an agent behaves in a production-like setting and decide whether the value justifies a larger commitment.
Those controls are particularly relevant for autonomous or semi-autonomous workflows. An agent that performs a multi-step task may invoke models repeatedly, call tools or work for longer than a simple chat interaction. Cost visibility should therefore be paired with technical observability: teams need to know which workflows drive consumption, whether the task succeeded and whether the same result could have been achieved with a smaller model, a shorter context or fewer steps. Billing controls limit exposure; they do not by themselves make a workflow efficient.
Google is also previewing deferred execution pricing for eligible workloads. The company says customers will be able to mark suitable agent work as deferred so an intelligent scheduler can run it during off-peak capacity windows. It states that eligible jobs could cost up to half as much and bypass standard quota limits. That option is described as coming soon and available for select workloads, so organisations should check current documentation before designing a process around it. It is best suited to work where elapsed time is less important than cost efficiency.
Developer access under one subscription
Another part of the release concerns developer tooling. Google says Antigravity in Gemini Enterprise and Android Studio AI use will be included in Gemini Enterprise subscriptions for eligible customers, with broader rollout planned. Usage across the app, platform and developer tools is intended to roll into a single view rather than separate licensing and billing silos. Centralised reporting can be useful when a business wants one owner to manage AI budgets while multiple technical teams use different interfaces.
The company says developer-tool quota will be pooled across the Google Cloud project. This is a practical detail with organisational consequences. It allows a shared capacity pool to support business applications, custom agents and coding work, but it also means teams need clear internal ownership. Administrators should decide who can permit overages, how cost centres are identified, what alerts reach project owners and which workloads are allowed to consume deferred capacity.
For organisations with sustained demand, Google has introduced Gemini Enterprise Flexible Savings Plans. The company says customers can commit to a monthly spend without a stated minimum or maximum and receive 10% savings for a one-year commitment or 20% for three years. The plans can draw down against an existing Google Cloud Enterprise Agreement. This is a different decision from enabling a product feature: it is a financial commitment that should follow a period of measured usage and a realistic view of likely demand.
What enterprises should evaluate next
The update gives Gemini Enterprise customers more levers, not a single best configuration. A sensible first step is to identify workloads by urgency and volatility. Everyday staff assistance may fit a seat model. A short-lived project or bursty agent may suit pay-as-you-go. A predictable background job could be a candidate for deferred execution if and when it becomes available. Those choices should then be tested with caps, attribution tags and alerts already enabled.
Teams should also keep availability language in view. Google describes several items as available only to select customers or rolling out broadly soon. Pricing, eligibility and regional access can change, especially for new product editions. The announcement is a useful signal of direction, but a procurement or engineering decision should use the current Gemini Enterprise and Google Cloud documentation that applies to the organisation’s account.
Google Cloud published the announcement on 26 August 2026. The pricing structures, cost controls and availability statements in this report are based on that official post. For organisations scaling agent workflows, the key takeaway is not simply lower cost: it is the ability to match billing, quota and guardrails more closely to the way different AI workloads actually run.