Skip to main content

Research

What Gemini in Looker Costs: Editions, User Licenses, and Data Tokens

What Looker's editions, user licenses, and Conversational Analytics data tokens cost, the token cost of one question at a pilot team's reported rate, and what changes when Google ends the promotion.

Reading time
4 min read
Sources
14 cited
In this piece
2 products

Key findings

  1. Looker (Google Cloud core) is sold through sales as an annual platform edition (Standard, Enterprise, or Embed) plus Developer, Standard, and Viewer licenses. Google publishes no list price, so ask for each license type priced on its own line of the quote.
  2. Each instance gets a monthly data token pool that does not roll over: 60M input and 1.2M output tokens on Standard, 300M and 6M on Enterprise, and 1.2B and 24M on Embed.
  3. Customers have promotional access with no quota limits or overage fees within fair use. Google will give at least 90 days' notice before overage of US$3.00 per million input and US$20.00 per million output tokens applies.
  4. Copilot in Power BI meters the same kind of work in Fabric capacity units: 100 CU seconds per 1,000 input tokens and 400 per 1,000 output tokens. It runs on paid F2 or P1 capacity and above.
  5. Of 65 public Looker accounts, 10 discuss cost and 9 raise it as a concern; none says Looker is not worth its price.

Measure Conversational Analytics token use now, while Google still waives overage. One pilot team reports about 250,000 input tokens per question, so a Standard instance's monthly input pool would cover roughly 240 questions for the whole company. The license is the fixed part of the bill, and warehouse compute grows with every answer.

Two meters on top of the license

Looker's license covers the platform and its users. Conversational Analytics adds a data token meter, counted per instance. Input tokens cover prompts, conversation history, LookML metadata, and agent instructions. Output tokens cover answers, generated queries, the reasoning shown in Thinking mode, and visualizations.

  • Looker charges an annual edition fee and user licenses through sales, and pools data tokens per instance with a monthly allocation set by the edition.
  • Copilot in Power BI needs Power BI Pro at US$14 or Premium Per User at US$24 per user per month, paid yearly, plus paid Fabric or Premium capacity. Copilot draws that capacity down in capacity units.
  • The warehouse bills every Looker dashboard tile, Explore, and agent answer as a query.

What one question costs in tokens

Input tokens drive the bill. An admin running a small pilot reported in September 2026 that each prompt consumed about 250,000 input tokens on agents built on single Explores of about 80 fields. The same admin's personal agent, with no custom instructions, used 75,000. Dashboard agents used noticeably more. A second admin in the thread explains that a dashboard agent carries the context of every Explore the dashboard reads, so a dashboard on 12 Explores sends about 12 times the context.

At that rate, the Standard pool of 60M covers about 240 questions a month and the Enterprise pool of 300M about 1,200, shared across the whole instance. Past the pool, each question's input would cost about US$0.75 at the announced overage rate.

Google's own product post puts each visible field at 40 to 50 tokens. Hiding unused fields, scoping agents to focused Explores, and keeping instructions short all cut input.

Measure during the promotion

Google calls the current terms promotional access, extended so organizations can baseline usage with its new observability tools. The Token usage tab of the Conversational Analytics dashboard in System Activity shows estimated tokens by day, starting from the day an admin enables tracking.

Measure tokens per active user and per agent now, then compare them with the edition's monthly pool before the 90-day notice arrives.

Teams report warehouse compute and license price

Warehouse compute is the cost Looker teams report first. After one company converted its Tableau dashboards to LookML dashboards, its IT team said they consumed too much compute, and users found filter changes slow. A data platform team at a 500-person company that has Claude build dashboards on Snowflake measured about 1 million tokens per dashboard, roughly US$0.50 each.

License price is the other complaint. A BI architect calls Looker pricier than almost all SaaS competitors, a team moving from Power BI found the license cost shocking on paper. A founder calls the price high for a small company. An agency that served 67 client companies on Looker for three years moved to Sigma at a similar cost. That same data platform team, citing a US$150,000 average Looker license, built its own dashboard host for about US$8,300 over eight months, and its Looker use fell by a third.

Methodology

GILD read Google's and Microsoft's official pricing and consumption pages on October 12, 2026, and two Google Developer forum threads on Conversational Analytics token use on October 25, 2026. Cost counts come from 65 public Looker accounts published from Feb 13, 2025 to Oct 26, 2026, each tagged by the topics it discusses. List prices are in US dollars and are the vendors' own. Token pools and overage rates are Google's published figures. The per-question token figures are one admin's pilot numbers, and the questions-per-pool and dollar estimates are GILD's arithmetic on them. The per-dashboard token figure is one data platform team's own measurement.

Products in this piece

Sources

Official documentation
7
Public source
7
  1. Looker pricingOfficial documentationcloud.google.com
  2. Looker docs: Looker (Google Cloud core) features and entitlementsOfficial documentationdocs.cloud.google.com
  3. Looker docs: Conversational Analytics in Looker overviewOfficial documentationdocs.cloud.google.com
  4. Looker release notesOfficial documentationdocs.cloud.google.com
  5. Microsoft Learn: Copilot in Fabric consumptionOfficial documentationlearn.microsoft.com
  6. Microsoft Learn: Copilot for Power BI overviewOfficial documentationlearn.microsoft.com
  7. Microsoft: Power BI pricingOfficial documentationmicrosoft.com
  8. Reddit post on dashboards converted to LookML consuming too much compute (Aug 11, 2026)Public sourcereddit.com
  9. Reddit comment on Looker's license cost next to Power BI (Apr 14, 2026)Public sourcereddit.com
  10. Reddit comment on moving from Looker to Sigma (Apr 16, 2026)Public sourcereddit.com
  11. G2 review from a BI architect on self-serve and price (Jun 27, 2025)Public sourceg2.com
  12. G2 review from a founder on LookML and price for a small company (Jul 25, 2025)Public sourceg2.com
  13. Google Developer forums: Conversational analytics / gemini token usage data (ruthc and others, Sep 14, 2026)Public sourcediscuss.google.dev
  14. Google Developer forums: Introducing Data Token Observability for Conversational Analytics in Looker (Tarunima, Aug 18, 2026)Public sourcediscuss.google.dev

Keep reading

  1. ComparisonLooker vs Microsoft Copilot: Gemini in Looker or Copilot in Power BI?Copilot in Power BI, part of Copilot in Fabric, bills Fabric capacity per token, and Gemini in Looker draws on a pooled data token allowance. The semantic model you already run should decide.
  2. ResearchBundled or Add-On: What Workspace and Microsoft 365 Charge for AI in 2026Google folds Gemini into Workspace plan prices while Microsoft sells Copilot as an add-on over free Copilot Chat. What each model costs per seat, and what it leaves out.
  3. ResearchWhat Looker Teams Report: Self-Serve, LookML, Cost, and the First AI AgentsWhat 65 public accounts from 2025 and 2026 say about Looker's semantic layer, self-serve, cost, charts, and AI agents built on LookML.
Read the research