specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
provider: Advanced Micro Devices
providerId: advanced-micro-devices
created: '2026-05-04'
# Provenance stamped 2026-08-11: this artifact was written by the API Evangelist
# bulk sweep dated 2026-05-04, not harvested from the provider. See roadmap#35.
method: generated
modified: '2026-05-05'
reconciled: false
notes: AMD-issued invoices typically reflect hardware/license purchases; recurring GPU-hour
  spend usually flows through cloud providers (AWS, Azure, OCI) rather than AMD directly.
tags:
  - FinOps
  - FOCUS
  - GPU
  - HPC
description: 'FOCUS-aligned FinOps for AMD: hardware purchase + cloud-provider-mediated GPU-hour
  consumption.'
sources:
  - https://www.amd.com/
  - https://rocm.docs.amd.com/
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: Advanced Micro Devices, Inc.
serviceCategory: Compute / Semiconductors
billingModel:
  pricingCategory: Pay-As-You-Go + Committed Use
  billingFrequency: Per-Invoice
  billingCurrency: USD
  chargeCategories:
    - Purchase
    - Usage
    - Adjustment
focusColumns:
  ServiceName: AMD
  ServiceCategory: Compute / Semiconductors
  ProviderName: Advanced Micro Devices
  PublisherName: Advanced Micro Devices, Inc.
  InvoiceIssuerName: Advanced Micro Devices, Inc.
  BillingCurrency: USD
meters:
  - name: gpu_hours
    unit: GPU-hour
    aggregation: sum
    dimensions:
      - sku
      - region
      - cloud_provider
  - name: hardware_units
    unit: unit
    aggregation: sum
    dimensions:
      - sku
      - reseller
  - name: support_subscriptions
    unit: month
    aggregation: sum
    dimensions:
      - support_tier
principles:
  - name: Visibility
    description: Track GPU-hour consumption through the hosting cloud provider's billing exports
      (CUR, OCI Usage, Azure Cost Management); attribute hardware spend via PO-level tagging.
  - name: Allocation
    description: Tag AMD-instance jobs by team/model/experiment in the cloud provider's billing
      data.
  - name: Optimization
    description: Use ROCm-native frameworks to maximize GPU utilization; right-size MI300X vs MI250X
      for workload; pursue committed-use discounts on the host cloud.
  - name: Accountability
    description: Assign GPU-hour budget owner; review monthly utilization and idle-GPU exposure.