Google Overhauls Gemini Enterprise Pricing, Companies Will Pay Only for Tokens Used

Google has announced an overhaul of Gemini Enterprise's pricing model, introducing usage-based billing in place of fixed subscriptions. The move responds to growing complaints from finance departments unable to predict how much running AI agents will cost their companies.

No More Paying for Unused Seats

Until now, companies using Gemini Enterprise had to buy subscriptions for a specific number of seats, regardless of how many were actually used by employees or AI agents. The new pay-as-you-go option removes that requirement entirely, charging only for tokens and model calls actually consumed, at standard API rates.

Google says the new billing option is currently rolling out to a select group of customers, with a broader rollout planned in the coming weeks. The company doesn't hide that this is a response to a specific industry problem: workloads generated by AI agents can vary drastically from month to month, making traditional budgeting practically impossible.

Discounts for Commitments and Deferred Tasks

For companies with more predictable workloads, Google has built Flexible Savings Plans, with discounts reaching 10 percent for a one-year commitment and 20 percent for a three-year one. Customers declare their own monthly spending amount, with no stated minimum or maximum, which is meant to give flexibility to both small and large deployments.

A separate new feature is Deferred Execution Pricing. Tasks that don't require immediate execution, such as model evaluations, document processing or data indexing, can be flagged to run during low-load hours on Google's infrastructure. In exchange, companies pay up to half as much for the inference itself. The feature currently covers a limited pool of tasks.

Hard Limits Instead of Bill Surprises

Google is also adding budget control tools directly into the Google Cloud billing console. Companies can set a hard monthly spending cap at the project level, and once it's reached, Gemini Enterprise agent activity is paused without affecting other Google Cloud workloads. Automatic email notifications go out at 50, 80 and 100 percent of the set amount.

On top of that come FinOps tools: anomaly detection for AI spending with root-cause analysis, an agent that generates natural-language cost summaries, and consolidated billing reports covering Gemini Enterprise, Google Antigravity and developer tools under a single subscription.

Pay-as-you-go lets you spin up an agent experiment on a Friday afternoon and pay only for what it actually consumes - Stephanie Walter, HyperFRAME Research

It's mostly a shift, not a discount. But matched to the right workload, it can genuinely save money - Manoj Chandra Jha, Nord-IQ Research

Competitive Pressure and Unpredictable Bills

The pricing shift is part of a broader fight for enterprise customers, in which Google is competing against Microsoft Copilot, OpenAI's ChatGPT Enterprise and Anthropic. All three players compete not just on per-token price but also on cost predictability, and a lack of control over AI agent spending has been one of the main barriers to large-scale deployments cited by IT and finance departments.

For Polish companies considering a Gemini Enterprise rollout, the new pricing means being able to test AI agents without signing annual contracts for a preset number of seats. That could lower the entry barrier especially for smaller IT teams, which previously had to estimate demand in advance, often paying for unused licenses.

Analysts note that flexible billing is less a price cut than a shift in financial risk, from a fixed preset budget to a payment model tied to actual usage. For companies with variable AI agent workloads, that could mean real savings, but it also requires new real-time spend-monitoring tools that many finance departments don't yet have.