Gemini Enterprise pricing and cost per seat across every edition, with overages, in 2026

Google's product page no longer shows a price. Until early October 2026 it stated two starting prices: Business "Starting at $21 USD per seat per month" and Standard / Plus "Starting at $30 USD per seat per month". Pay-as-you-go, for 20 or more seats, carried a "$0 seat fee" and billed only usage. Google never said whether the starting prices were annual or monthly, gave Plus no separate figure and quotes Frontline through sales. Secondary sources put Plus at $50 annual, with flex rates roughly 17 percent above annual.

The distinction matters more than it sounds. Because Google prorates daily rather than charging monthly in advance, adding or removing seats mid-cycle adjusts the bill proportionally instead of waiting for the next period. That is genuinely convenient. It also means the seat line on your invoice will rarely be a round number, which makes reconciling against a quoted rate harder than it should be.

Budget the seat as the floor rather than the price. The full model is edition rate times seats, plus consumption past the pooled quota, plus the internal cost of owning the Google Cloud project, its IAM roles, its data stores and its connector credentials.

Overages are consumption charges that apply once you exhaust your edition's included quota. Google publishes one rate directly: storage and data indexing at $5 per GiB per month, prorated by the duration you store the data. Overages for the other metered features, Assistant queries, image generation, video generation and Deep Research, are billed against Agent Platform prices rather than a rate published on the quotas page.

Two conditions gate them. You need an invoiced Cloud Billing account, and you need at least one active, non-free-trial subscription. Self-serve accounts on a trial cannot run an overage at all, which sounds protective and in practice means the feature simply stops working when the pool empties.

The storage line is the one to model first, because indexing is where enterprise content gets expensive. Standard includes 30 GiB pooled across the edition. A single mid-sized SharePoint or Drive corpus can exceed that on its own, and at $5 per GiB per month an extra 100 GiB is $500 a month, or $6,000 a year, on top of every seat you have already paid for.

There is no minimum on the mainstream editions: Google documents Standard, Plus and Pay-as-you-go as "1 or more users." The minimum that does exist is on Frontline, which requires "150 or more users of Standard or Plus editions" before you can buy it. Frontline is a companion tier for deskless staff, not an entry point, and the 2 GiB pooled storage reflects that.

The more binding constraint is a maximum rather than a minimum. Business is capped at "Up to 300 seats" on the product page (the editions docs still say "1-500 users"), and since August 19, 2026 Google has capped what you can transact through the Cloud console: 25 seats per self-serve or resold account, 1,000 per invoiced account. Those are purchasing limits, not product limits, but they land on the same calendar.

Because Google caps self-serve and resold billing accounts at 25 seats per account when subscriptions are managed through the Google Cloud console. Invoiced accounts are capped at 1,000. Google documents that if you need more, or if you hit the maximum seat limit error while adding licenses to an existing subscription, you contact your Google Cloud sales representative or account manager for an offline order, or request an allowlist exemption.

For a 40 or 60 person company this converts a self-serve purchase into a procurement cycle, with a sales conversation, an offline order and whatever lead time that carries. It is not a blocker and the exemption path exists, but it belongs in the project plan. Teams that assumed they could expand a pilot from 20 seats to 80 on a card are the ones this catches.

Quotas are expressed per user but consumed collectively. Google states that individual usage is not capped at the per-user limit and that all users in the edition draw from their edition's shared pool until it is depleted. Pools reset daily at midnight Pacific Time for every feature except the AI developer tools, and they are scoped per project and location.

The practical consequence is that a per-seat allowance is not a per-seat guarantee. On Standard, 100 seats buys 16,000 Assistant queries a day across the whole edition. If a dozen analysts run heavy sessions, they can consume a disproportionate share and the remaining 88 people hit a restricted feature through no fault of their own. Google is explicit that feature usage is restricted when you exceed the quotas of your edition.

This is not necessarily a bad design. Pooling is more forgiving than hard per-person caps for the common case where usage is uneven. It just means capacity planning has to be done on aggregate expected usage, not on headcount times the published per-user number, and the failure mode is a bad day for the light users rather than a bill.

The editions differ on pooled storage, daily feature quotas and seat eligibility rather than on search quality. Business allows up to 300 seats per the product page (the editions docs still say 1 to 500 users) with 25 GiB pooled. Standard and Plus both allow one or more users, with 30 GiB and 75 GiB respectively. Frontline carries only 2 GiB and requires 150 Standard or Plus seats to exist alongside it. Pay-as-you-go removes pooled license quotas entirely and bills consumption against an invoiced account.

The daily quotas track the same ladder. Standard allows 160 Assistant queries per user per day, 3 Deep Research runs, 5 images and 2 videos. Plus raises those to 200, 10, 10 and 3, and lifts no-code agent creation from 1 a day to 10. The Frontline tiers sit far below both, at 20 queries a day for Starter and 40 for Worker, which is a reasonable fit for shift workers and a poor one for analysts.

Choose on ceilings rather than on the price gap. The step from Standard to Plus buys 45 GiB of pooled storage and roughly triple the Deep Research allowance, and if either of those is your binding constraint the higher edition is usually cheaper than paying the overage on the lower one.

Four things: consumption past your pooled quota, any Google Cloud infrastructure the deployment touches, the procurement effort above 25 or 1,000 seats, and the internal engineering time to run it. Gemini Enterprise is a Google Cloud product rather than a Workspace add-on, so it needs a Cloud project and a billing account, and somebody has to own IAM roles, data store configuration, connector authentication and identity sync schedules.

That last line is the one that never appears in a comparison table and is frequently the largest at small and mid scale. There are also platform ceilings worth knowing before you design around them: a maximum of 500 data stores and 500 engines per project, 100 data stores and 10,000,000 documents per location, and 300 complete query requests per minute per project.