A team turns on an AI assistant, uses it hard for a week because it is new, and a month later somebody in accounting asks what a line on the Google Cloud invoice is. The tool is rarely the problem. The problem is that nobody decided in advance what too much looked like, and nothing in the account was configured to stop there.
Two entries in Google's Gemini Enterprise release notes this month change who gets to make that decision. They are worth reading closely, because the good news and the catch sit in the same sentence.
What changed, and when
On 1 September 2026, the Gemini Enterprise release notes recorded that configuring overage controls is available to all projects linked to an invoiced Cloud Billing account. Before that, the controls were limited to customers who had received a specific billing notification email.
On 10 September 2026, the same release notes recorded that subscribing to the Gemini Enterprise Pay-as-you-go edition, along with access to the AI developer tools, is also available to all projects linked to an invoiced Cloud Billing account. The earlier restriction to customers who had received that email no longer applies. Both entries checked on 15 September 2026.
Read as a headline, that is spending controls for everyone. Read as an administrator, it is narrower than it sounds.
Who this actually reaches
The phrase to read twice is invoiced Cloud Billing account. It is not a synonym for an account that gets a bill. It is a specific kind of Cloud Billing account, and Google's own page on applying for monthly invoiced billing lists the eligibility: you must be registered as a business for at least one year, you must expect to spend a minimum of USD 40,000 a year on Google Cloud, and invoiced billing has to be available in your country. Checked 15 September 2026.
Nearly every small business we onboard is on the other kind. You sign up, you put a card on the account, and Google charges that card on a monthly cycle. That is a self-serve billing account, and it is the right one for most companies under a few dozen people. It also means the two release notes above do not apply to you. You cannot enable overages and you cannot subscribe to the Pay-as-you-go edition, however much you would like the spend limit field that comes with them.
Here is the part that surprises people. On a self-serve account you are already protected, by accident. A Gemini Enterprise subscription entitles the project to a pooled feature quota. When the project reaches that pooled quota and overages are not enabled, feature usage stops and users see a usage limit message in the app. No overage means no runaway. The ceiling may be lower than you want, but it is a real ceiling and it is on by default.
What we check on a client project first
Before anyone touches a switch, we go through five things. It takes about twenty minutes and it prevents most of the billing surprises we get called about.
- Which billing account the project is linked to. Self-serve or invoiced. This single fact decides which half of this article applies to you.
- Which edition the subscription is on. Overage charges apply to the Standard, Plus and Standard Emerging Market editions. Other editions do not support overages at all, so the controls will not appear no matter how long you look for them.
- Where usage is visible today. Gemini Enterprise has a usage view in the Cloud console, and Cloud Billing has reports that break these costs out. If nobody on the client side knows where either one lives, a spend limit is theatre.
- Who holds the billing administrator role. Applying for invoiced billing requires a billing account administrator, and so does most of the cleanup afterwards. On small teams this is often one person, and sometimes that person has left.
- Whether a budget exists, and whether anyone reads the alerts. A budget alert routed to a shared inbox nobody opens is the same as no budget.
If you are on invoiced billing, set the limit before you enable the overage
Order matters here. Enabling overages first and configuring the limit afterwards leaves a window where usage is uncapped, and that window is exactly when a new tool gets hammered by a curious team.
- Confirm the prerequisites. Google's documentation states that overages require an invoiced Cloud Billing account, at least one active subscription that is not a free trial, and a supported edition.
- Set the project monthly spend limit. In the Google Cloud console, open Gemini Enterprise, go to the Usage and Spending page, select the Usage tab, and in the project monthly spend limit section, set the limit.
- Then enable overages. With the limit already in place, Gemini Enterprise stops overage usage automatically when the project reaches it.
- Back it with a Cloud Billing budget. Google's documentation for this flow says to configure your budget and spend cap actions in Cloud Billing, and when defining the budget scope, to select Vertex AI (aiplatform.googleapis.com) from the services list.
Google Cloud console
> Gemini Enterprise
> Usage and Spending
> Usage tab
> Project monthly spend limit > Set limit
One caveat comes straight from Google's own text and is worth repeating to whoever approves the spend: stopping usage can take a few minutes to take effect, so you might incur charges that exceed your limit. Set the number slightly under the figure you would actually be unhappy about, not exactly at it.
The budget alert trap
This one costs people real money and it has nothing to do with Gemini. Cloud Billing budgets and budget alerts do not cap your usage or your charges. Google's documentation says it plainly: alert emails might prompt you to take action to control your costs, but they do not automatically prevent the use or billing of your services when a threshold is met or exceeded. A budget alert is a smoke detector, not a sprinkler.
There is a separate feature, budget spend caps, which does pause the applicable service when a project exceeds the budget set for that service. As of our check on 15 September 2026, Google's documentation describes spend caps as available for select services only, so confirm that yours is covered rather than assuming it.
If the only thing between your client and a bad month is an alert email, say so in plain language. Most owners hear the word budget and assume the spending stops on its own.
Cost and risk notes
We are not going to print a per-seat figure for Gemini Enterprise here. Google's edition comparison points buyers of the Standard, Plus and Frontline editions to its sales team rather than publishing a rate, and a number typed from memory would be a guess dressed up as research. When you get a quote, get it per edition and per seat in writing, and ask as a separate question what usage above the included quota bills at. Those are two different numbers and only one of them is usually on the page you are handed.
The Pay-as-you-go edition deserves its own warning. Google describes it as a way to run agent workloads without hitting quota limits mid-task, with no upfront commitment and no base subscription fee, because you pay for what you use. The first half is the selling point. The second half is the risk: no feature quota limit also means no natural stopping point. The pooled quota that quietly protects a self-serve account is simply not there. On that edition a monthly spend limit is not a refinement you add later, it is part of turning the thing on.
Setting the limit too low has a cost as well, and it is the smaller one. Users hit a usage limit message in the middle of real work and the team loses an afternoon of goodwill toward the tool. We would rather that happen in week one, against a number somebody chose on purpose, than find the alternative on an invoice in month two.
What to watch in the first month
Put a recurring fifteen minutes in a calendar for the first four weeks. Open the Gemini Enterprise usage view and look at which features the quota is actually going to, because it is rarely the ones people predicted in the kickoff meeting. Open the Cloud Billing report filtered to the service and compare it against the limit you set. If usage is pressing against the ceiling every week, the real question is whether the edition is wrong, not whether the limit is wrong.
Then write down what you decided and why, somewhere the next person will find it. A spend limit with no note attached gets raised by whoever is annoyed first, and nobody remembers what the original number meant.
How Guanacos Tech helps
We are an independent consultancy with Google-certified engineers, and this is ordinary work for us: reading which billing account a project is really on, checking whether the controls a release note describes are even available to that account, setting the limit before the switch rather than after it, and leaving the client with a short page that says what to watch and who to call. No rebuild, no drama.
If you want someone to look at your project before you turn an AI tool loose on it, see what we do on Google Cloud or book a call.
Sources
- Gemini Enterprise release notes (Google Cloud)
- Overview of overages and spend controls (Gemini Enterprise)
- Configure overages (Gemini Enterprise)
- Compare editions of Gemini Enterprise
- Apply for monthly invoiced billing (Cloud Billing)
- Create, edit, or delete budgets and budget alerts (Cloud Billing)