specification: API Commons Plans
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/Plans
provider: CUDO Compute
providerId: cudo-compute
created: '2026-06-21'
modified: '2026-06-21'
reconciled: false
tags:
- GPU
- Cloud Compute
- Infrastructure
- Virtual Machines
- Marketplace
- Plans
description: >-
  CUDO Compute uses a usage-based, pay-as-you-go model billed per GPU-hour, per
  vCPU-hour, per GB of memory-hour, and per GB of storage, with rates that vary
  by GPU model and data center. On-demand VMs are billed by the second/hour;
  Reserved Capacity and committed contracts unlock lower rates across multiple
  term lengths. Bare metal and multi-node GPU clusters are typically priced by
  quote/reservation. The API exposes live machine-type prices per data center.
notes: >-
  CUDO Compute does not publish a single fixed public per-GPU-hour rate card
  for every model; on-demand rates are surfaced per data center and GPU model
  in the console and via the machine-type-prices endpoint, and newer/enterprise
  GPUs are quote-on-request. The per-GPU-hour figures below are representative
  examples drawn from secondary pricing comparisons and CUDO blog/product pages
  and MUST be reconciled against the live catalog
  (GET /v1/data-centers/{dataCenterId}/machine-type-prices) and
  https://www.cudocompute.com/pricing before use.
sources:
- https://www.cudocompute.com/pricing
- https://www.cudocompute.com/docs/api
- https://www.cudocompute.com/products/gpu-cloud/nvidia-a100
- https://www.cudocompute.com/blog/what-does-it-cost-to-rent-cloud-gpus
plans:
- id: cudo-compute-on-demand
  name: On-Demand
  type: usage
  description: >-
    Pay-as-you-go GPU and CPU virtual machines billed per resource-hour with no
    commitment; rates vary by GPU model and data center.
  entries:
  - label: NVIDIA H100 (per GPU)
    name: gpu_h100
    type: usage
    metric: gpu_hours
    limit: -1
    timeFrame: hour
    geo: global
    unit: 1
    price: ~$1.80-$2.45 per GPU/hr (representative; varies by data center)
    userMultiplied: false
  - label: NVIDIA H200 (per GPU)
    name: gpu_h200
    type: usage
    metric: gpu_hours
    limit: -1
    timeFrame: hour
    geo: global
    unit: 1
    price: quote on request / per data center
    userMultiplied: false
  - label: NVIDIA A100 80GB (per GPU)
    name: gpu_a100
    type: usage
    metric: gpu_hours
    limit: -1
    timeFrame: hour
    geo: global
    unit: 1
    price: ~$1.50-$1.60 per GPU/hr (representative; varies by data center)
    userMultiplied: false
  - label: NVIDIA L40S (per GPU)
    name: gpu_l40s
    type: usage
    metric: gpu_hours
    limit: -1
    timeFrame: hour
    geo: global
    unit: 1
    price: per data center (see live catalog)
    userMultiplied: false
  - label: NVIDIA V100 (per GPU)
    name: gpu_v100
    type: usage
    metric: gpu_hours
    limit: -1
    timeFrame: hour
    geo: global
    unit: 1
    price: per data center (see live catalog)
    userMultiplied: false
  - label: NVIDIA A800 (per GPU)
    name: gpu_a800
    type: usage
    metric: gpu_hours
    limit: -1
    timeFrame: hour
    geo: global
    unit: 1
    price: per data center (see live catalog)
    userMultiplied: false
  - label: vCPU
    name: vcpu
    type: usage
    metric: vcpu_hours
    limit: -1
    timeFrame: hour
    geo: global
    unit: 1
    price: per vCPU/hr (varies by CPU model and data center)
    userMultiplied: false
  - label: Memory
    name: memory
    type: usage
    metric: gib_hours
    limit: -1
    timeFrame: hour
    geo: global
    unit: 1
    price: per GiB/hr (varies by data center)
    userMultiplied: false
  - label: Block Storage
    name: storage
    type: usage
    metric: gib_month
    limit: -1
    timeFrame: month
    geo: global
    unit: 1
    price: per GiB/month (varies by storage class and data center)
    userMultiplied: false
  elements:
  - name: Virtual Machines
  - name: Disks & Storage
  - name: Networks
  - name: Public IPs
- id: cudo-compute-reserved
  name: Reserved Capacity
  type: subscription
  description: >-
    Commitment-based capacity at discounted rates across multiple term lengths,
    with capacity guarantees for GPU VMs, dedicated servers, and clusters.
  entries:
  - label: Committed GPU Capacity
    name: reserved_gpu
    type: subscription
    metric: gpu_hours
    limit: -1
    timeFrame: month
    geo: global
    unit: 1
    price: discounted vs. on-demand (term-based; contact sales)
    userMultiplied: false
  elements:
  - name: Reserved GPU VMs
  - name: Commitment Terms
- id: cudo-compute-bare-metal-cluster
  name: Bare Metal & Clusters
  type: enterprise
  description: >-
    Dedicated bare metal servers and multi-node GPU clusters (H200, H100,
    A100 80GB, L40S, A800, and Blackwell-generation GPUs) priced by quote and
    reservation with InfiniBand/high-bandwidth networking options.
  entries:
  - label: Cluster Reservation
    name: cluster_quote
    type: flat
    metric: contract
    limit: -1
    timeFrame: month
    geo: global
    unit: 1
    price: quote on request
    userMultiplied: false
  elements:
  - name: Bare Metal Servers
  - name: Multi-Node GPU Clusters
  - name: High-Bandwidth Networking
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com