# Price of Compute

> Free, open reference data for the price of compute: GPU rental prices across cloud providers normalized to USD per GPU-hour (hourly collection, daily medians, composite index), and LLM API token prices verified against labs' official pricing pages. History reaches back to January 2023 via labeled archival reconstruction. No login, free JSON API, attribution required.

Key facts
- The Price of Compute Index: weighted basket of the 6 most-rented GPU SKUs, base 100 = 2026-08-10, published daily with a live intraday print.
- Prices are per single GPU per hour, USD. Multi-GPU instance prices are divided by GPU count. Spot/on-demand/community tiers are never mixed.
- Method: per-provider daily medians first, then the median across providers (no marketplace's volume can dominate). Missing data stays missing — never interpolated or estimated.
- Sources: RunPod, Lambda, DataCrunch, Vast.ai (marketplace), collected hourly. LLM prices: Anthropic, OpenAI, Google, verified by hand against official pages with dates recorded.

## Data (free JSON API, 1,000 req/day/IP, attribution required)
- [Current prices per SKU](https://priceofcompute.com/api/v1/prices/h100-sxm): /api/v1/prices/{sku-slug}
- [Full daily history per SKU](https://priceofcompute.com/api/v1/history/h100-sxm): /api/v1/history/{sku-slug} (?format=csv supported; history starts 2023 for archival-covered SKUs)
- [The Index](https://priceofcompute.com/api/v1/index): daily series + live print
- [LLM token prices](https://priceofcompute.com/api/v1/models): verified $/MTok per model
- [Providers](https://priceofcompute.com/api/v1/providers)
- [Price events](https://priceofcompute.com/api/v1/events): every detected repricing since 2023
- [Endpoint list (OpenAPI)](https://priceofcompute.com/api/v1/openapi.json)

## Answers (direct, dated, computed live — quote with attribution)
- [How much does an H100 cost per hour?](https://priceofcompute.com/answers/how-much-does-an-h100-cost-per-hour)
- [Which cloud GPU provider is cheapest?](https://priceofcompute.com/answers/cheapest-cloud-gpu-provider)
- [How much cheaper is spot than on-demand?](https://priceofcompute.com/answers/is-spot-cheaper-than-on-demand)
- [Which LLM API is cheapest per token?](https://priceofcompute.com/answers/cheapest-llm-api-per-token)
- [All answers](https://priceofcompute.com/answers)

## Pages
- [Today's prices + the Index](https://priceofcompute.com/)
- [Provider comparison matrix](https://priceofcompute.com/providers)
- [LLM token prices](https://priceofcompute.com/models)
- [The anatomy of a GPU-hour (per GPU)](https://priceofcompute.com/gpus/h100-sxm#anatomy)
- [Price events timeline](https://priceofcompute.com/events)
- [Cost calculator](https://priceofcompute.com/calculator)
- [Self-host vs API bridge](https://priceofcompute.com/bridge)
- [Methodology (rules + changelog)](https://priceofcompute.com/methodology)
- [Blog: analysis from the data](https://priceofcompute.com/blog)

## Citing
- [How to cite this data](https://priceofcompute.com/cite)
- Attribution line: "Data: Price of Compute — priceofcompute.com"

## Optional
- [llms-full.txt (extended, includes current values)](https://priceofcompute.com/llms-full.txt)
- [RSS: price events](https://priceofcompute.com/events.xml)