Google AI Enterprise Pricing: 2026 Seat & Token Costs

Google's enterprise AI pricing splits into two products with different meters. Gemini Enterprise (which replaced the old Gemini for Google Workspace add-ons in October 2025) runs $21 per seat per month for the Business tier and $30 per seat per month for the Standard/Plus tier, both on annual commitments, with a $18/seat/month Frontline tier for deskless workers and a free Gemini Enterprise Starter tier capped at limited usage. Vertex AI — the developer platform — is priced per million tokens: Gemini 3 Pro runs roughly $2.00 per million input tokens and $12.00 per million output tokens, while Gemini 2.5 Flash sits near $0.30 input / $2.50 output. Google Cloud requires a 12-month minimum on Gemini Enterprise seats. Most US service businesses under 50 employees land between $3,800 and $18,000 per year all-in.

The Seat Pricing, Broken Down Honestly

Google restructured this in late 2025 and the old pricing pages still floating around are wrong. Here's the current shape:

The number that surprises people: this is on top of Google Workspace, not instead of it. A Workspace Business Standard seat is $14/user/month annual. So a 25-person HVAC company running Workspace Business Standard plus Gemini Enterprise Business is at $35/user/month, or $10,500 per year. That's the real line item, not $21.

Vertex AI: The Meter That Actually Bites

If you're building anything — a booking agent, a lead qualifier, a call summarizer — you're on Vertex AI, and seat pricing is irrelevant. Token pricing as of late 2025:

Real math from a plumbing client: 4,200 inbound calls per month, transcribed and summarized, roughly 3,500 input tokens and 400 output tokens each. On Flash that's $5.60/month. On 3 Pro it's $49.56/month. The model choice matters more than 10x, and most service-business tasks — classify this lead, summarize this call, draft this follow-up — do not need the frontier model. We route roughly 80% of production traffic to Flash-tier models and reserve Pro for genuinely hard reasoning.

There's also Provisioned Throughput, Google's reserved-capacity product. It's sold in GSU (Generative AI Scale Units) on 1-week, 1-month, or 1-year terms. A single GSU month for Gemini Flash runs in the low hundreds of dollars and guarantees a fixed throughput floor. Below roughly $8,000/month in on-demand spend, it almost never pencils out.

The seat price is what you negotiate. The token bill is what you actually pay. Companies that budget only for seats blow their forecast in month three.

Where the Hidden Costs Live

Six line items that don't appear on Google's pricing page:

When Google Enterprise AI Is Not Worth It — Read This Part

This section costs us business, and it's the most useful thing on the page.

Do not buy Gemini Enterprise seats if you have under 15 employees. At 8 people you're spending $2,016/year for seat licenses that mostly duplicate what a $20/month consumer Gemini or ChatGPT subscription does. Buy three individual subscriptions for the people who'll actually use them and put the other $1,500 into ads.

Do not buy it if you're not already on Google Workspace. The entire value proposition is grounding on your Gmail, Drive, and Calendar data. On Microsoft 365, Copilot at $30/user/month will beat it for the same reason — proximity to your data. Switching your whole office suite to justify an AI add-on is a $40,000 mistake dressed as a $10,000 decision.

Do not buy Provisioned Throughput before you have 90 days of usage data. We've seen a client commit to a 1-year GSU term and utilize under 20% of reserved capacity. That's real money burned for a latency guarantee they didn't need.

Do not expect it to fix a broken sales process. If your team takes 4 hours to return a lead call, an AI that drafts a better email is decorating the problem. Speed-to-lead is the variable; the median service business we audit responds in over 3 hours, and the ones who get under 5 minutes see conversion improve dramatically regardless of what model wrote the message.

Known failure modes, plainly:

Who should buy it: organizations with 25+ Workspace seats, a real document corpus, someone accountable for adoption, and a specific workflow — not "we should do AI." If you can't name the workflow in one sentence, you're not ready.

What It Actually Costs to Get Value

A realistic first-year budget for a 30-person US service business:

That needs to return more than roughly $45,000 in margin to be defensible. For a business doing $3M with a 12% net margin, that's about 15 additional closed jobs at a $2,500 average ticket, assuming the AI is genuinely responsible for them. Run that math before you sign. Our ROI calculator does it with your numbers, and our pricing page shows what implementation actually costs when someone else builds it.

How to Buy It Without Overpaying

The honest summary: Google's enterprise AI is priced fairly and the models are excellent. The failure is almost never the technology — it's buying 60 seats for a 45-person company, using 14 of them, and having no one own the outcome. Pick one workflow, meter it, and expand only where the numbers hold.