The GPU Compute Index

Name: The GPU Compute Index (MCI): a volume-weighted, normalised USD-per-GPU-hour price index for hourly GPU rentals, refreshed every 5 minutes from API integrations plus web scraping across 50+ providers (30+ surfaced in the index), with 15 tracked GPU types each carrying its own provider count and 1-day/7-day/30-day change series, selectable historical windows (7/30/60/90-day and all-time), sub-indices by GPU tier and pricing type, and adjacent research on hardware depreciation and resale value.
Creator: Mercatus AI

MERCATUS AIA5 · direct accessManage supplier listing

The GPU Compute Index (MCI): a volume-weighted, normalised USD-per-GPU-hour price index for hourly GPU rentals, refreshed every 5 minutes from API integrations plus web scraping across 50+ providers (30+ surfaced in the index), with 15 tracked GPU types each carrying its own provider count and 1-day/7-day/30-day change series, selectable historical windows (7/30/60/90-day and all-time), sub-indices by GPU tier and pricing type, and adjacent research on hardware depreciation and resale value.

Open buyer room →

Sample

Brickroad customers

Sample this dataset before you buy it.

Your sourcing agent asks Mercatus AI and files the sample in your catalog — private to you.

Start 7-day trialNothing is charged until the trial ends.

Coverage

Universe, instruments and categories.

Industry: Financial Exchanges & DataIT Consulting & Other Services
Instruments: equitiesoptions
Categories: PriceAlternative
Regions: US
Sample tickers: NVDAAMDCRWVNBISIRENAPLDMSFTGOOGLAMZNSMCI

Dataset card

Type: not stated
Format: Tabular (API/JSON index series with per-GPU detail pages and selectable historical windows)
Volume: 15 GPU types × 3-26 providers each (H100 26, H200 19, A100 80GB 17, L40S 15, B200 11, B300 7, MI300X 6) at 5-minute collection cadence with a daily-published index; a 90-day rolling trend series per model plus an all-time window, and per-model 1d/7d/30d change series — roughly 7K provider-model observations per day at the stated cadence if the full underlying set is retained
Users: No consumer registration base is published — access is key-based through a self-serve developer dashboard, so the meaningful count is active API keys rather than registrants, and the operating company records 2 employees.
History: not stated
Update frequency: not stated
Growth: Active — index updated daily with period-high dated four days before review, 5-minute underlying collection cadence, an expanding per-GPU detail-page set, and an adjacent research stream (H100 depreciation, resale value, and an advocacy piece arguing for open price indices in AI compute) indicating an actively developed product
Launched: est. 2025-2026 (no launch date published). Bounded by an 'ALL' historical window control existing alongside a 90-day default series, and by the organisation record of 2 employees with no funding history — consistent with a recent small build rather than an established index house
Delivery: not stated
Entity mapping: not stated
Sample: not stated
Point-in-time: not stated
Licence: not stated

supplier index read from their site

Mercatus AI has not added their own details yet. Not yet on file:

Dataset
Data dictionary
Coverage
Provenance
Rights & data handling