specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
provider: Union.ai
providerId: unionai
created: '2026-06-20'
modified: '2026-06-20'
reconciled: false
tags:
- AI
- ML
- Orchestration
- Workflows
- Serverless
- FinOps
- Cost Management
- FOCUS
description: >-
  FinOps view of Union.ai spend. Union bills usage-based pricing for the managed
  orchestration plus the action-allocated compute (CPU, Memory, GPU) each
  workflow consumes, metered per second and never for idle capacity. The headline
  meters are actions and per-second vCPU-hours, GB-hours, and GPU-hours. The
  open-source Flyte engine itself is free (Apache-2.0); only the managed Union
  platform and the underlying cloud compute incur charges. Enterprise / BYOC
  terms are negotiated and largely unpublished, so this artifact is not
  reconciled.
notes: >-
  Per-second resource rates and action pricing change; verify against the
  Union.ai pricing page during reconciliation. In BYOC deployments the raw
  cloud-compute cost lands on the customer's own cloud bill, separate from the
  Union platform fee - allocate both when modeling total cost.
sources:
- https://www.union.ai/pricing
- https://www.union.ai/blog-post/introducing-union-serverless
- 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: Union.ai
serviceCategory: AI and Machine Learning
billingModel:
  pricingCategory: Usage-Based
  billingFrequency: Monthly
  billingCurrency: USD
  chargeCategories:
  - Usage
  - Purchase
  - Adjustment
focusColumns:
  ServiceName: Union Serverless
  ServiceCategory: AI and Machine Learning
  ProviderName: Union.ai
  PublisherName: Union.ai
  InvoiceIssuerName: Union.ai
  BillingCurrency: USD
  ChargeCategory: Usage
  PricingCategory: Usage-Based
meters:
- name: actions
  description: Task actions executed by workflows, billed per action.
  unit: actions
  aggregation: sum
  dimensions:
  - tenant
  - project
  - domain
  - workflow
- name: vcpu_hours
  description: Allocated vCPU time across action containers, metered per second.
  unit: vcpu_hours
  aggregation: sum
  dimensions:
  - tenant
  - project
  - domain
  - task
- name: memory_gb_hours
  description: Allocated memory across action containers, metered per second.
  unit: gb_hours
  aggregation: sum
  dimensions:
  - tenant
  - project
  - domain
  - task
- name: gpu_hours
  description: Allocated GPU accelerator time (T4g through B200), metered per second.
  unit: gpu_hours
  aggregation: sum
  dimensions:
  - tenant
  - project
  - domain
  - task
  - gpu_type
- name: platform_fee
  description: Fixed monthly Team-plan platform fee (includes an applied usage credit).
  unit: subscription
  aggregation: sum
  dimensions:
  - tenant
principles:
- name: Visibility
  description: Pull Union usage reporting per project/domain/workflow; in BYOC also reconcile raw cloud-compute cost on your own cloud bill.
- name: Allocation
  description: Tag projects and domains to cost centers; allocate action and resource-hour spend per workflow/team.
- name: Optimization
  description: Right-size task requests (CPU/Memory/GPU), pick the smallest sufficient GPU, use Actors to amortize warm containers, and avoid over-allocating resources that bill per second.
- name: Accountability
  description: Assign owners per project; review action counts and GPU-hour burn monthly against budget; negotiate Enterprise volume discounts at scale.
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com