specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
provider: Google Quantum AI
providerId: google-quantum-ai
created: '2026-05-25'
modified: '2026-05-25'
reconciled: true
tags:
  - FinOps
  - Cost Management
  - Quantum Computing
  - Sponsored Access
description: |
  FinOps profile for Google Quantum AI. Unlike most Google Cloud services, the Quantum Engine API today
  exposes no metered SKU on the Cloud Billing surface. Cost management therefore centres on:
    - Sponsorship and approved-user lists (capacity-as-grant rather than pay-as-you-go).
    - Reservation grants and budgets that allocate processor minutes/hours to a project.
    - Open-source software costs (none — Apache 2.0).
    - Indirect costs (engineering time, classical simulator compute for qsim, GPU hours for qsim/cuQuantum
      and TensorFlow Quantum workloads).
sources:
  - https://quantumai.google/cirq/google/concepts
  - https://quantumai.google/cirq/google/access
  - https://quantumai.google/willowearlyaccess
  - https://cloud.google.com/billing/docs
alignedWith:
  framework: FinOps Foundation Framework
  frameworkUrl: https://www.finops.org/framework/
  dataSpec: FOCUS
  dataSpecVersion: '1.3'
  dataSpecUrl: https://focus.finops.org/focus-specification/v1-3/
publisherName: Google LLC
serviceCategory: Quantum Computing
billingModel:
  pricingCategory: Sponsored / Grant-Based
  billingFrequency: Not Billed
  billingCurrency: USD
  chargeCategories:
    - Usage (informational only — no charge)
  primaryUnit: processor-minute
  secondaryUnits:
    - job
    - shot
    - qubit-second
costAllocation:
  recommendedDimensions:
    - projectId
    - processorId
    - reservationGrantId
    - jobLabels.team
    - jobLabels.experiment
  description: Allocate costs (even if zero) by tagging programs and jobs with the team and experiment
    labels exposed by the Quantum Engine API. Labels propagate to job records and surface in
    `listjobs` filters.
budgeting:
  recommendedControls:
    - Track total processor-minutes consumed per reservation grant.
    - Track open-swim minutes used per user via job-event timestamps.
    - Track classical simulator GPU hours separately (these *do* incur Cloud Compute charges).
optimisation:
  recommendations:
    - Use the willow_pink Quantum Virtual Machine (QVM) and qsim locally to debug before submitting hardware
      jobs.
    - Batch sweep parameters into a single job to amortise queueing overhead and stay under the 5-minute
      open-swim cap.
    - Schedule heavy workloads inside reservation windows; reserve open-swim for short interactive runs.
    - Avoid leaving idle reservations on the calendar — cancel via :cancel to return time to the pool.
reporting:
  recommendedReports:
    - Processor-minutes consumed per reservation grant per week.
    - Open-swim minutes per approved user per month.
    - Job success rate by processor and by calibration window.
    - Classical simulator (qsim / cuQuantum) GPU hours per experiment.