specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: microsoft-azure-cost-management
providerId: microsoft-azure-cost-management
created: '2026-05-04'
modified: '2026-09-17'
generated: '2026-09-17'
method: searched
source: https://learn.microsoft.com/en-us/azure/cost-management-billing/costs/manage-automation
tags:
  - Rate Limiting
  - Quotas
  - Throttling
  - FinOps
description: >-
  Published Cost Management limits, replacing the 2026-05-04 scaffold that carried invented
  X-RateLimit-* headers and invented per-tier quotas. Cost Management does NOT use the generic
  X-RateLimit-* family. It throttles on two independent mechanisms: the Azure Resource Manager
  platform limits that apply to every ARM call, and — unique to this resource provider — a
  Query Processing Unit (QPU) budget that prices a call by how much data it asks for rather than
  by how many calls you make. The QPU quotas are per TENANT, not per subscription or per key,
  so every team in an organisation shares one budget.
headers:
  qpuConsumed: x-ms-ratelimit-microsoft.costmanagement-qpu-consumed
  qpuRemaining: x-ms-ratelimit-microsoft.costmanagement-qpu-remaining
  qpuRetryAfter: x-ms-ratelimit-microsoft.costmanagement-qpu-retry-after
  consumptionRetryAfter: x-ms-ratelimit-microsoft.consumption-retry-after
  retryAfter: Retry-After
headers_note: >-
  Read x-ms-ratelimit-microsoft.costmanagement-qpu-remaining on every response and slow down
  before you are throttled — it is a list of the remaining quotas, one per window. On a 429, the
  number of seconds to wait is in the QPU retry-after header on Query, and in
  x-ms-ratelimit-microsoft.consumption-retry-after or retry-after on the report generators.
  Sources: the manage-automation doc for the QPU headers, and the ErrorResponse /
  GenerateCostDetailsReportErrorResponse schema descriptions in the 2026-06-01 contract for the
  retry headers.
responseCodes:
  throttled: 429
  serviceUnavailable: 503
  payloadTooLarge: 413
limits:
  - name: Query API — QPU per 10 seconds
    scope: tenant
    metric: query-processing-units
    limit: 12
    window: 10s
    applies_to:
      - Query_Usage
      - Query_UsageByExternalCloudProviderType
    source: https://learn.microsoft.com/en-us/azure/cost-management-billing/costs/manage-automation
  - name: Query API — QPU per minute
    scope: tenant
    metric: query-processing-units
    limit: 60
    window: 1m
    applies_to:
      - Query_Usage
      - Query_UsageByExternalCloudProviderType
    source: https://learn.microsoft.com/en-us/azure/cost-management-billing/costs/manage-automation
  - name: Query API — QPU per hour
    scope: tenant
    metric: query-processing-units
    limit: 600
    window: 1h
    applies_to:
      - Query_Usage
      - Query_UsageByExternalCloudProviderType
    source: https://learn.microsoft.com/en-us/azure/cost-management-billing/costs/manage-automation
qpu_model:
  unit: Query Processing Unit
  definition: >-
    A performance currency abstracting CPU and memory, described by Microsoft as analogous to
    Cosmos DB RUs. One QPU is currently deducted per MONTH OF DATA queried.
  drivers:
    - Date range — the wider the requested window, the more QPU the call consumes.
  volatility: >-
    Microsoft states the QPU calculation logic and the quotas may change without notice, and
    that further QPU factors may be added. Do not hard-code the 12/60/600 numbers into a
    scheduler; read the remaining header.
  source: https://learn.microsoft.com/en-us/azure/cost-management-billing/costs/manage-automation
guidance:
  recommended_call_frequency: no more than once per day per scope
  reason: Cost data refreshes every four hours; more frequent calls return the same data and
    only add load.
  report_generation: Generate a cost details report no more than once a day for a given scope
    and date range; split large pulls into daily or weekly windows.
  large_result_sets: >-
    GenerateDetailedCostReport returns 413 when the request would exceed 2 GB — move to Exports
    rather than retrying.
platform_limits:
  note: >-
    Cost Management is an ARM resource provider, so the Azure Resource Manager per-subscription
    and per-tenant read/write throttling applies underneath the QPU model. Those limits are
    documented at the platform layer and are not Cost-Management-specific.
  source: https://learn.microsoft.com/en-us/azure/azure-resource-manager/management/request-limits-and-throttling
limit_count: 3