ai-automation

Gemini for Business: Pricing and Billing Explained (2026)

Automation Architects Team·5 November 2026·8 min read
Gemini for Business: Pricing and Billing Explained (2026)

Gemini business pricing confuses people because "Gemini" is three separate purchases with three separate bills. There's Gemini inside Google Workspace, paid per user with your Workspace licence. There's Gemini Enterprise, Google's agent app, paid per seat. And there's the Gemini API on Google Cloud, paid per million tokens. This guide takes each in turn: what it costs, when you're charged, the traps, and the South African payment and VAT details.

If you're still deciding whether Gemini is the right platform at all, start with our Gemini platform overview. For the same guide on Anthropic's side, see Claude Team billing explained.

Which Gemini bill are you actually paying?

What you buy Priced Billed through
Gemini in Google Workspace Per user per month, inside the Workspace plan Your Workspace subscription
Gemini Enterprise (agent app) Per seat per month, or pay-as-you-go Google Cloud
Gemini API / Agent Platform Per million tokens in and out A Google Cloud billing account

Most mid-sized businesses start with the first and only reach the third when they automate a process. Knowing which line you're on avoids the classic mistake of buying a second AI licence for capability you already pay for.

How much does Gemini in Google Workspace cost?

Gemini is now included in every Workspace Business plan, at different depths. These are Google's published US-dollar prices per user per month:

Plan Annual plan Flexible plan Gemini included
Business Starter USD 7 USD 8.40 Gemini in Gmail
Business Standard USD 14 USD 16.80 Plus Docs, Sheets, Slides, Meet and Chat
Business Plus USD 22 USD 26.40 As Standard, with stronger security and archiving

Business plans are capped at 300 users. Higher tiers can add an AI Expanded Access add-on for higher limits, priced by local currency and term. (Source, Source)

Before buying anything, check which tier you're on. Because Google folded Gemini into the plans, plenty of businesses already pay for it and haven't switched it on.

Flexible or Annual: which Workspace plan should you choose?

This is where Workspace bills surprise finance teams.

  • Flexible Plan. Add and remove users whenever you like. You're billed monthly in arrears, prorated by day for the users you had. It costs about 20% more per user.
  • Annual/Fixed-Term Plan. You commit to a licence count for the year at the lower price. You can add licences, prorated, but you can't reduce them until renewal, and cancelling early means paying the full commitment. Payment can still be monthly. (Source)

It's the same trap as on most seat-based AI plans: someone leaves, the licence frees up, and the bill doesn't move. Go annual for the stable core of the team and keep fast-changing groups, such as contractors and seasonal staff, in mind when you set the count.

What does Gemini Enterprise cost?

Gemini Enterprise is Google's agent app: one place to search company data and run agents across Workspace, Microsoft 365 and other systems. It's billed per seat per month and needs an active user licence.

  • The Business edition starts at USD 21 per seat per month, with a 30-day free trial, for teams of up to 300.
  • Standard and Plus editions have no public price. They're quoted by Google Cloud sales, with higher quotas and unlimited seats. (Source, Source)

Since August 2026 you can also mix seats with a pay-as-you-go edition, billed at standard model API rates with no upfront commitment. Quotas pool across users in a project, admins decide whether overage is allowed, and Flexible Savings Plans take 10% off a one-year or 20% off a three-year monthly spend commitment. (Source)

How is the Gemini API billed?

Anything you build on Gemini Enterprise Agent Platform (the evolution of Vertex AI) or the Gemini API is pay-as-you-go per million tokens, on a Google Cloud billing account. Prices are listed in US dollars. Here are the two models we use most, on Google's standard tier and global endpoint:

Model Input Cached input Output
Gemini 4 Argon (from 1 Dec 2026) USD 4.00 USD 0.20 USD 20.00
Gemini 3.8 Flash (until 31 Dec 2026) USD 0.75 USD 0.075 USD 3.75

Figures are per million tokens. (Source)

Five details that change the bill:

  • The introductory price arrives as a credit. Until 30 November 2026, Google returns 50% of net spend on Gemini 4 Argon as credits, issued on the 7th of the following month and expiring 30 days later. Your invoice shows the full rate. The effective price is USD 2 in and USD 10 out. Gemini 3.8 Flash has the same arrangement until 31 December 2026, after which it's USD 1.50 in and USD 7.50 out. (Source)
  • Batch is half price. Work that can wait a few hours costs 50% of the standard rate.
  • Regional endpoints cost about 10% more than the global endpoint.
  • Long answers can bill even when they fail. From 13 November 2026, Gemini 4 completions over 32K tokens are charged for retrievable tokens even if the request times out. Long-running generation should stream, or be split.
  • Grounding with Google Search is free for 5,000 queries a month across Gemini 3 and 4 models, then USD 14 per 1,000.

How do you stop a Gemini API bill running away?

Google Cloud spend caps became generally available on 30 October 2026, and they're the control we switch on first. For AI services a cap triggers within minutes using estimated costs, resets automatically each month, and alerts at 50%, 80% and 100% by default. One exception: Provisioned Throughput is never halted by a cap. (Source)

Budgets and alerts on their own only send email. A cap stops spending. On a new agent, set both.

What should South African buyers watch?

  • Workspace is paid by card, in USD or EUR. Google's payment-options page lists credit or debit card for South Africa and doesn't list monthly invoicing. Use a card without forex fees, because every monthly charge and prorated licence goes through it. (Source)
  • VAT applies. Google says certain transactions by foreign suppliers to South African customers are subject to South African VAT, and you can add your VAT ID to invoices. Agree the treatment with your accountant before the first invoice. (Source)
  • Cloud is a separate account. API and Agent Platform usage bills to a Google Cloud billing account, not to your Workspace subscription. Decide early who owns it, and put the spend cap on it.
  • Your data and the model are in different places. Google Cloud's Johannesburg region can hold your data, but Gemini models aren't offered there, so the model call crosses the border. That's a POPIA design question, not a pricing one, and our Gemini overview explains how we handle it.

A one-page checklist for finance

  • Check your current Workspace tier before buying any Gemini add-on.
  • Put the stable core of the team on the Annual plan and remember licences can't come down mid-term.
  • Decide whether you need Gemini Enterprise at all, or whether Workspace plus one well-built agent does the job.
  • For the API, budget at the standard rate, not the introductory one, and expect promotional credits on the following month's statement.
  • Turn on a spend cap on every project that calls a model.
  • Use a card without forex fees, add your VAT ID, and agree the VAT treatment with your accountant.

Get the licence right, then get the value

The licence is the cheap part. The return comes from the processes Gemini takes over: inference running inside BigQuery, documents classified on arrival, agents grounded in your own data. Our Gemini implementation service covers licence planning and spend controls through to production agents on Google Cloud, and because we're model-agnostic we'll tell you when a workload belongs on Claude, GPT or Copilot instead.

Want to know what Gemini should cost for your team? Book a Free AI Assessment and we'll price it against your actual tier and volumes.

GeminiGoogle WorkspaceGemini EnterpriseAI PricingBillingSouth Africa

Next step

Want to know what this would look like in your business?

Our free AI assessment is a scoped conversation about your systems, your constraints and what is actually worth automating — not a product demo. You leave with a plan you can act on, whether or not you go on to work with us.

Get a Free AI Assessment

Related posts