San Francisco Compute Company logo

San Francisco Compute Company

San Francisco Compute Company (SF Compute) operates a market for buying short-duration time on H100 and H200 GPU clusters with no long-term contracts. Customers reserve VM nodes for any quantity, duration, and start time through the sf CLI and the sfcompute.com Orders API, with managed Slurm and bare-metal clusters available by request.

1 APIs 6 Features
AICloudClustersComputeGPUMachine LearningMarketplaceTraining

San Francisco Compute Company publishes 1 API on the APIs.io network. Tagged areas include AI, Cloud, Clusters, Compute, and GPU.

San Francisco Compute Company’s developer surface includes documentation, signup flow, pricing, engineering blog, changelog, support, CLI, and 9 more developer resources.

APIs

SF Compute Orders API

The SF Compute Orders API is the REST control plane for placing, listing, and managing orders that reserve GPU node-hours on H100 and H200 clusters. Orders specify node count, d...

Features

Market-Based GPU Pricing

Per-node hourly pricing fluctuates with supply and demand, with no long-term contracts.

Flexible Node Reservations

Reserve any number of H100 or H200 nodes for custom durations and start times.

Managed Slurm Clusters

Slurm-scheduled clusters available on request for larger training jobs.

Bare Metal Clusters

Direct bare-metal access for advanced workloads on request.

Hardware Failure Refunds

Automatic refunds for failed nodes with hardware-level monitoring.

Cancel Anytime

Cancel or resell unused capacity without penalty.

Resources

🔗
Website
Website
🔗
Developer
Developer
🔗
Documentation
Documentation
📝
SignUp
SignUp
🔗
Login
Login
💰
Pricing
Pricing
📰
Blog
Blog
📄
ChangeLog
ChangeLog
💬
Support
Support
👥
GitHubOrganization
GitHubOrganization
🔗
LinkedIn
LinkedIn
🔗
CLI
CLI
📦
SDK
SDK
📦
SDK
SDK
📦
SDK
SDK
📦
SDK
SDK

Sources

apis.yml Raw ↑
aid: sf-compute
name: San Francisco Compute Company
description: San Francisco Compute Company (SF Compute) operates a market for buying short-duration time
  on H100 and H200 GPU clusters with no long-term contracts. Customers reserve VM nodes for any quantity,
  duration, and start time through the sf CLI and the sfcompute.com Orders API, with managed Slurm and bare-metal
  clusters available by request.
image: https://kinlane-productions.s3.amazonaws.com/apis-json/apis-json-logo.jpg
url: https://raw.githubusercontent.com/api-evangelist/sf-compute/refs/heads/main/apis.yml
created: '2026-05-23'
modified: '2026-05-23'
specificationVersion: '0.19'
type: Index
access: 3rd-Party
position: Producer
tags:
- AI
- Cloud
- Clusters
- Compute
- GPU
- Machine Learning
- Marketplace
- Training
apis:
- aid: sf-compute:orders-api
  name: SF Compute Orders API
  description: The SF Compute Orders API is the REST control plane for placing, listing, and managing orders
    that reserve GPU node-hours on H100 and H200 clusters. Orders specify node count, duration, and start
    time. Authentication uses bearer tokens generated via the sf CLI.
  humanURL: https://docs.sfcompute.com
  baseURL: https://api.sfcompute.com/v0
  tags:
  - Clusters
  - GPU
  - H100
  - H200
  - Orders
  - REST
  properties:
  - type: Documentation
    url: https://docs.sfcompute.com
common:
- type: Website
  url: https://sfcompute.com
- type: Developer
  url: https://docs.sfcompute.com
- type: Documentation
  url: https://docs.sfcompute.com
- type: SignUp
  url: https://sfcompute.com/signup
- type: Login
  url: https://sfcompute.com/signin
- type: Pricing
  url: https://sfcompute.com/prices
- type: Blog
  url: https://sfcompute.com/blog
- type: ChangeLog
  url: https://sfcompute.com/changelog
- type: Support
  url: https://sfcompute.com/contact
- type: GitHubOrganization
  url: https://github.com/sfcompute
- type: LinkedIn
  url: https://www.linkedin.com/company/sf-compute
- name: sf CLI
  url: https://github.com/sfcompute/cli
  type: CLI
- name: Nodes API Go SDK
  url: https://github.com/sfcompute/nodes-go
  type: SDK
- name: Nodes API TypeScript SDK
  url: https://github.com/sfcompute/nodes-typescript
  type: SDK
- name: SFC Go SDK
  url: https://github.com/sfcompute/sfc-go
  type: SDK
- name: SFC TypeScript SDK
  url: https://github.com/sfcompute/sfc-sdk-typescript
  type: SDK
- type: Features
  data:
  - name: Market-Based GPU Pricing
    description: Per-node hourly pricing fluctuates with supply and demand, with no long-term contracts.
  - name: Flexible Node Reservations
    description: Reserve any number of H100 or H200 nodes for custom durations and start times.
  - name: Managed Slurm Clusters
    description: Slurm-scheduled clusters available on request for larger training jobs.
  - name: Bare Metal Clusters
    description: Direct bare-metal access for advanced workloads on request.
  - name: Hardware Failure Refunds
    description: Automatic refunds for failed nodes with hardware-level monitoring.
  - name: Cancel Anytime
    description: Cancel or resell unused capacity without penalty.
- type: GPUs
  data:
  - name: NVIDIA H100
    description: 80GB Hopper GPU nodes available with market-based pricing (around $1.97/hr average).
  - name: NVIDIA H200
    description: 141GB HBM3e Hopper GPU nodes available for hourly and longer-term rental.
  - name: NVIDIA B300
    description: Blackwell Ultra GPU nodes listed as upcoming on the SF Compute roadmap.
- type: Customers
  data:
  - name: NVIDIA
  - name: Roboflow
  - name: Datology
  - name: Liquid AI
  - name: MIT
maintainers:
- FN: Kin Lane
  email: [email protected]
  url: https://apievangelist.com