Home  /  Research Videos  /  Google
Google · 4:28 · Buyer-side briefing

The Consumption Side: Tokens, Agents, Avatars, and the Commit

Session 4 of the Google AI Negotiation Series. How agents, avatars, and media are sold per unit: Vertex tokens and the January 2027 reset, agents that bill every reasoning pass, video per second, the Gemini Enterprise commit plans with their symmetric downside, the Google Cloud commitment, and the three year illustration line by line.

Share

The presenters in this briefing are AI generated avatars. The research, figures, and guidance are real, produced by Redress Compliance analysts from our consulting engagements and market network.

Transcript

Full narration of the briefing. Click a section heading to jump the player to that moment.

Tokens and the January reset 0:00

Session four: the consumption side, where agents, avatars, and media are sold per unit. Vertex AI and the Gemini API bill per million tokens, input and output priced separately, reasoning tokens billed as output. The newest Flash generation carries introductory rates until December thirty first, twenty twenty six, and roughly doubles from January first, twenty twenty seven. Context caching, batch processing, and provisioned throughput each carry their own discount or premium.

I am Claire, Tom is with me, and this is the Google AI Negotiation Series. The January reset is the largest silent change in this note.

Agents, avatars, and media 0:45

Agents next. Prebuilt agents inside Gemini Enterprise draw on the seat quota and then on credits. Custom agents built with the agent development kit or Agent Builder bill as tokens, tool calls, grounding requests, and, where they run continuously, compute. A production agent that runs several reasoning passes per task bills the output rate on every pass, so an estimate taken at the demo loop arrives open.

Then avatars and video: AI presenters in Vids and the Gemini app, and video generation through Veo, metered per second of output. A minute of generated video costs more than a day of text.

Commit plans and the Cloud commitment 1:30

Two more mechanisms. Gemini Enterprise commit plans: since August twenty twenty six, consumption can be bought pay as you go, on a Flexible Savings Plan, or on a committed plan. The plans discount the rate in exchange for a floor, and the downside is symmetric: a plan sized to Google's forecast that covers less usage than expected costs more than paying as you go. And the Google Cloud commitment itself: a multi year committed spend agreement with tiered discounting, committed use discounts on compute, GPUs, and TPUs, and Marketplace drawdown subject to caps.

Since June twenty twenty six the default CUD scope is the billing account.

Where the money sits: the meters 2:12

Where the consumption side money sits, on the note's illustration: five thousand knowledge workers, a Gemini Enterprise rollout, two production agents, and a media program, over three years at approximate list. Vertex tokens forecast on introductory rates: three point six million; with rates held for the term, a volume tier from month one, and batch and cached rates specified: two point one. Production agents estimated at the demo loop: one point four million or open; forecast on the production loop with tool calls and grounding priced and a monthly ceiling: nine hundred thousand. Avatars, video, and image: unforecast; with one media ceiling, alerts, and an approval gate: four hundred thousand.

Where the money sits: the two big lines 3:01

The two big lines. The Gemini Enterprise commit plan: Google's structure is a three year committed plan sized to Google's forecast, nine million dollars against six point seven million of usage. The buyer side structure is a Flexible Savings Plan at eighty percent of the buyer's forecast with the remainder pay as you go: six point nine million. And the Google Cloud commit: a thirty six million dollar flat annual floor, against twenty seven million ramped, with shortfall rollover, CUD scope confirmed, and Marketplace drawdown to the cap.

The total: fifty million plus open lines, against thirty seven point three million, all lines capped.

Four lessons 3:41

Four lessons before the response. The commit is the money: a flat floor sized to Google's model against a ramped floor sized to yours dwarfs every discount. The January reset is the largest silent change: usage that does not move at all costs twice as much in year two unless the rate is in the paper. And open lines are the surprise: media, avatars, and agent tool calls are individually modest and collectively the difference between a budget and an invoice.

The full research note is free under this video, at redresscompliance dot com slash google ai negotiation series.

The research playbook behind this briefing

Google AI in 2026: The Seat Side, the Consumption Side, and the Commit That Joins Them

This briefing is drawn from the full playbook by Redress Compliance: the preparation runway, the estate math, the give and get table, the tactics and counters, and the concessions checklist. Read it here, save the PDF, or send it to whoever owns the renewal.

PDF, free, no form. Opens in the page on desktop, or in your browser's own viewer on a phone.

Negotiating a Google renewal this year?

Redress Compliance works on contingency: our fee is 25 percent of what we save you. Nothing saved, nothing paid. Independent, buyer side only, never vendor funded. Want Redress to contact you? Reach out and we respond the same day.

Talk to a Google negotiator
Browse all 217 research videos