
Compute Screener tracks what AI compute costs.
Example path
https://computescreener.com/mcpCompute Screener: free compute prices. It publishes two daily price indices, for GPU rental and for AI tokens, and each exchange's compute futures prices. GPU prices are recalculated every minute (each source read every 1 to 30 minutes); token prices once a day. GPU rental: USD per GPU-hour for H100, H200, A100, B200, B300, MI300X, RTX PRO 6000, each standardized to an on-demand, 8-GPU, US contract; one value per day in US Central time, the current day provisional. AI tokens: USD per 1M tokens (3 input : 1 output blend) for Anthropic, OpenAI, Google, xAI, Meta, Moonshot AI, Alibaba, DeepSeek, Z.ai, Xiaomi, one value per UTC day. Changes (change_1d_pct, change_1w_pct, change_1m_pct) are percent changes from the value that many days earlier. Each GPU and company has a url: its page with the chart, history and a description. The figures are published under CC BY 4.0 with credit to Compute Screener (terms: https://computescreener.com/terms). They are information about posted prices and markets. Methodology: https://computescreener.com/methodology. Tools: (the market in one answer), and (latest indices), (RTX 5090, RTX 4090, L40S and 30 more GPUs as posted), (daily history), (a price at a past time), (tokens per company per day), (offers, in stock, futures volume), (futures against the index), (break-even throughput and cost per 1M tokens on each GPU), (which GPUs fit a model and the cost), (each AI model's posted API prices), (each GPU provider's posted prices), and (what a workload would cost), (ranges and volatility), (what moved), (everything covered), (a final value's manifest and how anyone can recompute it from the public files), (freshness and sources), search and fetch (the site's pages).
None in the store listings or on the server