specification: FinOps Framework
specificationVersion: '1.0'
schema: https://www.finops.org/framework/
provider: Cerebrium
providerId: cerebrium
created: '2026-06-20'
modified: '2026-06-20'
reconciled: true
tags:
- AI
- GPU
- Serverless
- Inference
- ML Infrastructure
- FinOps
- Cost Management
- FOCUS
description: >-
  FinOps view of Cerebrium spend. Cerebrium bills compute by the second:
  GPU time at a per-second rate per accelerator type (T4 through B200), CPU per
  vCPU-second, and memory per GB-second, charged only while a function executes.
  Persistent storage is billed per GB-month (first 100GB free). Flat platform
  fees apply at the Standard tier ($100/month) on top of compute; Hobby is free
  plus compute; Enterprise is custom.
notes: >-
  Per-second GPU/CPU/memory rates and the GPU catalog are subject to change;
  verify against the Cerebrium pricing page during reconciliation.
sources:
- https://www.cerebrium.ai/pricing
- https://www.cerebrium.ai/docs
- 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: Cerebrium
serviceCategory: AI and Machine Learning
billingModel:
  pricingCategory: Usage-Based
  billingFrequency: Monthly
  billingCurrency: USD
  chargeCategories:
  - Usage
  - Purchase
  - Adjustment
focusColumns:
  ServiceName: Cerebrium Serverless GPU
  ServiceCategory: AI and Machine Learning
  ProviderName: Cerebrium
  PublisherName: Cerebrium
  InvoiceIssuerName: Cerebrium
  BillingCurrency: USD
  ChargeCategory: Usage
  PricingCategory: Usage-Based
meters:
- name: gpu_seconds
  description: GPU runtime billed per second per accelerator type (T4, L4, A10, L40s, A100, RTX PRO 6000, H100, H200, B200).
  unit: seconds
  aggregation: sum
  dimensions:
  - account
  - app
  - gpu_type
- name: vcpu_seconds
  description: CPU runtime billed per vCPU per second ($0.00000655/vCPU/s).
  unit: seconds
  aggregation: sum
  dimensions:
  - account
  - app
- name: memory_gb_seconds
  description: Memory billed per GB per second ($0.00000222/GB/s).
  unit: gb_seconds
  aggregation: sum
  dimensions:
  - account
  - app
- name: storage_gb_month
  description: Persistent volume storage billed per GB-month ($0.05/GB/month; first 100GB free).
  unit: gb_month
  aggregation: sum
  dimensions:
  - account
  - app
- name: platform_fee
  description: Flat monthly platform fee at the Standard tier ($100/month); $0 for Hobby.
  unit: subscription
  aggregation: sum
  dimensions:
  - account
  - plan
principles:
- name: Visibility
  description: Pull Cerebrium usage and billing data; inspect per-app GPU-second burn by accelerator type.
- name: Allocation
  description: Tag apps per workload/team; map per-second compute to internal cost centers.
- name: Optimization
  description: Right-size GPU type to the workload (use T4/L4/A10 where sufficient); enable scale-to-zero; tune min/max replicas and cold-start storage; cut idle time since billing is per second.
- name: Accountability
  description: Assign owners per app; review per-second compute spend monthly against budget.
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com