API Toolkit · Documentation
 
Invitation only
API Toolkit

Programmatic Access
to ICX Reference Rates

Embed independently calculated GPU compute reference rates directly into your systems. Simple REST API, JSON responses, available for qualified subscribers.

Current Rates — Live Historical Rates — Live Webhooks — Coming Soon Forward Curves — Planned

Introduction

The ICX API provides programmatic access to independently calculated reference rates for GPU compute. Rates are derived from a curated panel of cloud providers using a transparent trimmed-mean methodology — the same methodology published on the public site.

All API responses return JSON. Rates are denominated in USD per GPU-hour. The API is versioned and backwards-compatible within a major version.

Available Indexes

The following indexes are currently live and accessible via API:

ICX H100S
H100 SXM5 80GB
NVLink interconnect · Min panel: 8
ICX H100P
H100 PCIe 80GB
PCIe interconnect · Min panel: 3

Authentication

Authenticate using a Bearer token passed in the Authorization header. API keys are issued per subscriber and scoped to specific indexes and rate limits.

Example Request Header
Authorization: Bearer icx_live_xxxxxxxxxxxxxxxxxxxx

API keys are available to qualified subscribers. Use the waitlist form below to request access. Do not share your API key or expose it in client-side code.

Base URL

https://icx.global/api/v1

All endpoints documented below are relative to this base URL. The current public beta uses unversioned endpoints (https://icx.global/api/) which are available for internal use. The versioned /v1 path will be available at GA.

Current Rates

Returns the latest computed ICX rates and the full provider panel that produced them.

GET /api/v1/rates Current ICX rates

Query Parameters

ParameterTypeRequiredDescription
gpu_type string optional Filter to a specific GPU type. Omit to return all active indexes. See GPU Types for valid values.

Response

200 OK
{
  "ok": true,
  "rates": {
    "H100 SXM5 80GB": {
      "rate": 2.87,
      "panelSize": 14,
      "providerCount": 10,
      "computedAt": "2026-04-01T12:00:00Z"
    },
    "H100 PCIe 80GB": {
      "rate": 1.94,
      "panelSize": 6,
      "providerCount": 5,
      "computedAt": "2026-04-01T12:00:00Z"
    }
  },
  "fetchedAt": "2026-04-01T12:34:56Z"
}

Code Example

const resp = await fetch('https://icx.global/api/v1/rates', {
  headers: { 'Authorization': 'Bearer YOUR_API_KEY' }
});
const { rates } = await resp.json();
console.log(`ICX H100 SXM5: $` + rates['H100 SXM5 80GB'].rate);
import requests

resp = requests.get(
    'https://icx.global/api/v1/rates',
    headers={'Authorization': 'Bearer YOUR_API_KEY'}
)
rates = resp.json()['rates']
print(f"ICX H100 SXM5: ${rates['H100 SXM5 80GB']['rate']}")
curl -H "Authorization: Bearer YOUR_API_KEY" \
  https://icx.global/api/v1/rates

Rate History

Returns a time series of ICX rates for the requested GPU type. Useful for charting, volatility analysis, or anchoring forward negotiations to observed price trends.

GET /api/v1/history Historical ICX rate series

Query Parameters

ParameterTypeRequiredDescription
gpu_type string optional GPU type to return history for. Defaults to H100 SXM5 80GB.
days integer optional Number of days of history. Default: 90. Max: 365 for subscribers.

Response

200 OK
{
  "ok": true,
  "gpu_type": "H100 SXM5 80GB",
  "history": [
    {
      "computed_at": "2026-01-01T12:00:00Z",
      "rate": 2.91,
      "min": 2.10,
      "max": 4.20,
      "providerCount": 12
    },
    // ... additional entries
  ]
}

All Indexes

Returns metadata on all tracked indexes — both live and announced-but-not-yet-live.

GET /api/v1/indexes Index catalog
200 OK
{
  "ok": true,
  "indexes": [
    {
      "id": "ICX-H100S",
      "gpu_type": "H100 SXM5 80GB",
      "label": "ICX H100 SXM5",
      "status": "live",
      "min_panel": 8,
      "update_frequency": "6h"
    },
    {
      "id": "ICX-H100P",
      "gpu_type": "H100 PCIe 80GB",
      "label": "ICX H100 PCIe",
      "status": "live",
      "min_panel": 3,
      "update_frequency": "6h"
    },
    {
      "id": "ICX-A100",
      "gpu_type": "A100 SXM4 80GB",
      "label": "ICX A100",
      "status": "coming_soon"
    }
  ]
}

GPU Types

Use these exact string values in the gpu_type query parameter:

ValueIndexStatus
H100 SXM5 80GBICX H100SLive
H100 PCIe 80GBICX H100PLive
A100 SXM4 80GBICX A100Coming Soon
H200 SXM5 141GBICX H200Planned

Methodology

ICX rates are computed using a trimmed arithmetic mean applied to a curated panel of cloud GPU providers:

  • Panel assembled from available on-demand H100 offerings, normalized to $/GPU-hour.
  • Minimum panel size enforced: 8 providers for H100 SXM5, 3 for H100 PCIe. Rate withheld if panel is insufficient.
  • Bottom and top 15% of prices trimmed before averaging.
  • Three-layer validation applied: bounds check, delta check (vs prior reading), IQR outlier detection.
  • Rates updated every 6 hours from the Get Deploying API. Each computation is stored with the full provider snapshot for audit.

The full methodology is documented on the public site.

Error Codes

HTTP StatusCodeMeaning
200ok: trueSuccess
401unauthorizedMissing or invalid API key
429rate_limitedToo many requests. See your plan's rate limit.
500internal_errorServer error. Check error field for details.
503panel_insufficientMinimum panel size not met — rate cannot be published.

Request API Access

The ICX API is currently in private beta. Access is available to qualified organizations. Complete the form below and we'll be in touch within 2 business days.

Join the API Waitlist

Tell us about your use case. Priority access for procurement teams, trading desks, and infrastructure platforms embedding ICX in production workflows.

✓ Request received — we'll be in touch within 2 business days.