specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
provider: TensorFlow
providerId: tensorflow
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: true
tags:
  - FinOps
  - FOCUS
  - Machine Learning
  - Open Source
description: >-
  FOCUS-aligned FinOps for TensorFlow: TensorFlow itself is free open-source software (Apache 2.0); the
  FinOps surface is the underlying compute, accelerator, and storage spend incurred by training and serving
  TensorFlow workloads on cloud or on-prem infrastructure.
sources:
  - https://www.tensorflow.org/
  - https://github.com/tensorflow/tensorflow/blob/master/LICENSE
  - https://focus.finops.org/focus-specification/v1-3/
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: TensorFlow contributors / Google LLC (project sponsor)
serviceCategory: Machine Learning Framework
billingModel:
  pricingCategory: Open Source
  billingFrequency: On-Demand
  billingCurrency: USD
  chargeCategories:
    - Usage
focusColumns:
  ServiceName: TensorFlow
  ServiceCategory: Machine Learning Framework
  ProviderName: TensorFlow
  PublisherName: TensorFlow contributors / Google LLC
  BillingCurrency: USD
meters:
  - name: training_accelerator_hours
    unit: instance-hour
    aggregation: sum
    dimensions:
      - accelerator_type
      - region
  - name: inference_accelerator_hours
    unit: instance-hour
    aggregation: sum
    dimensions:
      - accelerator_type
      - region
  - name: model_storage
    unit: GB-month
    aggregation: avg
    dimensions:
      - storage_class
principles:
  - name: Visibility
    description: TensorFlow itself emits no billing telemetry; visibility comes from the underlying cloud
      provider's cost / usage reports (e.g. GCP Billing export, AWS CUR) and from MLOps tracking (TensorBoard
      run metadata, TFX pipeline metrics).
  - name: Allocation
    description: Tag training and inference jobs with project / experiment / model_id labels at the cloud
      resource level so FOCUS BillingAccountId / tags can attribute spend back to the model owner.
  - name: Optimization
    description: Reduce TensorFlow infrastructure cost via mixed-precision training, distillation / quantization
      for inference, batch and concurrency tuning in TensorFlow Serving, spot/preemptible accelerators
      for training, and selective use of TPUs vs GPUs for the workload shape.
  - name: Accountability
    description: Model owner / ML platform team is accountable for per-model training and serving spend
      and reviews unit cost per prediction or per training run on a regular cadence.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com