MARKET PULSEH100 $4.18/GPU-HR 3PH200 $4.47/GPU-HR 1PMI300X $2.59/GPU-HR 1PL40S $1.62/GPU-HR 2PB300 $16.0/GPU-HR 1POBSERVED 2026-10-20 · 1H CACHE

FIELD GUIDE /

AI Agent Guide to GPU Prices and Compute Market Data

Retrieve SiliconTicker GPU rental listings, compute indices and official news through public JSON APIs. Preserve source, units, freshness and citation context.

Public compute data for AI agents

SiliconTicker publishes GPU rental observations, a public-cloud reference index and source-linked infrastructure news. Public read endpoints require no authentication. Use the human guide for context, then retrieve JSON for numerical values.

Begin with the short retrieval guide, the complete field guide and the OpenAPI 3.1 contract. These describe the supported data interfaces and their limitations.

Choose the right endpoint

Public, read-only JSON endpoints
EndpointUseSchemaInterpretation
/api/compareRental offers and plan costs1.0camelCase offer fields; availability is unconfirmed
/api/marketsPublic price summaries and CL-FCR2.0List-price references; not trades or executable quotes
/api/markets/azure-spotRegional Azure Spot tariffs1.0Capacity-contingent; effective date differs from observation date
/api/newsOriginal analysis and official releases2.1Publication age and source retrieval health are separate
/api/sourcesSource provenance, rights and cadence2.0Respect each source’s redistribution terms
/api/healthConnector and freshness status2.0HTTP 200 alone does not establish new data

Retrieve, validate and keep the clocks separate

  1. Read source health and choose the endpoint matching the question.
  2. Check schema_version and the endpoint’s freshness fields. Preserve stale or fallback labels.
  3. Read the observation timestamp before citing a value. checked_at is a retrieval clock, not proof of a new market observation.
  4. Preserve the unit and source. Divide plan totals only by the reported GPU count; do not invent missing fields.
  5. Confirm availability with the provider before treating any listing as bookable capacity.
GET https://siliconticker.com/api/compare
Accept: application/json

# Read offers[].gpu, gpuCount, totalHourlyUsd,
# perGpuHourlyUsd, pricingType, sourceUrl, observedAt,
# status and availability before comparing costs.
# Check the schema in /openapi.json for response structure.

The comparison catalog uses camelCase fields. The market and Azure APIs use snake_case fields. Consult the contract for each endpoint instead of carrying field names from one response to another.

Cite a price with its context

A useful citation includes SiliconTicker, the canonical page or API URL, upstream provider or dataset, numerical value, currency, unit, GPU variant, plan size, pricing type, region, observation date and freshness status. Include the upstream attribution and license where required.

CITATION TEMPLATE · NO PRICE IMPLIED

“SiliconTicker reports [value] USD per GPU-hour for [GPU variant] in [region], from [source], observed [date], status [fresh/stale/fallback]. This is a [public listing/Spot tariff]; availability is unconfirmed.” Link to the matching API record and provider source.

For CL-FCR, cite the canonical index page and preserve its constituent coverage and tradeable=false flag. For news, distinguish SiliconTicker analysis from the external publisher’s release; preserve the publication timestamp and canonical URL.

What the data does and does not establish

  • Prices: advertised rates and vendor tariffs do not establish executed trades, guaranteed capacity or a full workload bill.
  • Index: CL-FCR is a current-snapshot public-list median, not an investable product or a derivatives settlement benchmark.
  • Futures: read the dated venue status in the market roadmap and latest CME briefing; do not infer launch from a calendar date.
  • Rights: upstream restrictions still apply. ORNN values remain gated until redistribution rights are documented.
  • Failures: examine the freshness and provenance fields even after a successful request. Never replace unavailable observations with invented rates.

For workload comparisons, read GPU rental pricing, H100 versus A100 and Spot versus on-demand. The RSS feed carries editorial updates and source-linked releases.

Primary sources

Technical and billing details were reviewed on . Provider terms and capacity can change; price tables show their own observation dates.

Compare GPU rental offers →