Loading...
Daily snapshots derived from observable marketplace data. We publish what we see — nothing more, nothing less.
These aren't guidelines. They're constraints built into our system.
If data is sparse, inconsistent, or insufficient, we publish nothing. An empty cell is a signal — not a failure. We never interpolate or estimate.
We publish deterministic daily snapshots once every active marketplace collector has completed for the UTC day. Partial days are never published. 'Latest' means the most recent complete collection.
We report what we observe: listing prices, availability counts, sample sizes. Every price is read from its source on the day it is recorded, never from a stored or fixed price list. We never publish predictions, recommendations, or profitability projections.
Daily prices on this site are AIMC reference prices. Each authorized marketplace contributes one figure, the median of its own listings for that GPU that day, and the published price is the median across marketplaces. Every marketplace counts once, whatever its listing count, so no source can move the price by listing volume.
Each listing from an authorized partner is recorded at its published per-GPU hourly price, with no inference, smoothing, or interpolation. Each marketplace then contributes one median, so a source that lists the same price in many regions still counts once. Prices are in USD per GPU-hour: a source’s own USD list where it publishes one, otherwise converted at collection with that day’s ECB reference rate, with no fallback rate. Missing data shows as an em dash, never an estimate.
The daily price for each GPU is the median of the per-marketplace medians. When the middle-ranked marketplace reprices, the price moves to the next marketplace in the ranking, never further, so one source cannot move it arbitrarily.
Listings above 10× that marketplace's own 7-day median for the GPU are filtered as mispriced outliers; beyond that, prices are published exactly as observed — a representative market rate from raw listing snapshots, not a smoothed or modeled index.
Catalog-priced GPUs — chips with stable pricing across the marketplaces that carry them — may appear flat in time series. This reflects real market behavior, not a flaw in the chart.
The AIMC Market Index is a 0–100 daily score measuring GPU rental market conditions across five pricing tiers (on-demand, spot, reserved, community, hyperscaler). Built entirely from observable marketplace data — no predictions, no sentiment, no estimates.
Each tier is scored independently using only the listings that belong to it:
Each tier's score is built from six observational components measuring:
The headline Index score is the simple average of the tier scores available that day (normally all five):
Equal weights are chosen for transparency. A tier with no data that day is left out rather than estimated, and a component whose input lacks sufficient history is scored at zero weight with the remaining weights rescaled; the weights actually applied are stored on every record. Per-tier scores are always published alongside the composite.
Methodology v3 became effective 2026-06-02, adding the hyperscaler tier (public reference pricing from 10 major cloud providers). Methodology v2 (effective 2026-05-31) introduced 4-tier composite scoring across on-demand, spot, reserved, and community. Index values from 2026-03-13 through 2026-05-30 use v1, which scored only on-demand listings. The index has daily values since 2026-03-13 and has been computed on its own day since launch on 2026-03-25. All three methodologies remain queryable in the underlying database — every Index record stores its methodology_version for full audit traceability.
2026-11-08. We completed a review of every source collector against our published rules. Observations that were not a posted on-demand price for a whole, dedicated GPU of the stated SKU are now excluded from the on-demand tier: committed-term prices (now in the reserved tier of their stated term), shared or partitioned GPUs, aggregate means rather than posted offers, and SKUs whose form factor the source does not state (now published under the bare name). No observation was deleted; each carries a dated exclusion reason. Reference prices on this site are recomputed from the remaining observations and are higher for several SKUs as a result; this reflects the correction, not a market move. Published Index values are not revised.
We either have the data or we don't. If we have it, we publish it with full precision. If we don't, we show an em dash — no estimates, no gap-fills.
Data available from authorized marketplace partners. Published with full precision and listing counts visible.
Insufficient data to publish responsibly. Shows as empty — a deliberate signal.
We operate as an independent price-data publisher. Our governance rests on independence from the markets we price, a versioned and openly published methodology, documented conflict-of-interest constraints, and a formal process for queries and complaints — each described below.
Ai Mining Co. holds no positions in GPUs or compute derivatives, owns no GPU marketplace or trading venue, and earns no affiliate or referral revenue from the marketplaces it tracks. Our published figures are independent of any party with a financial interest in the prices we publish.
Every methodology revision is versioned, dated, and recorded. Each published record stores its methodology version for full audit traceability, and prior versions remain queryable. We do not revise published Index values retroactively. Reference-price history is recomputed from the current observation set; every correction that changes it is dated on this page.
When input data is insufficient to publish a determination responsibly, we publish nothing — shown as an em dash — rather than estimate or interpolate. Absence is a documented signal, not a gap to be filled.
Stakeholders who rely on our published figures may raise questions or formally challenge a published figure by contacting lucas@aiminingco.com. We review every submission, respond directly, and document any resulting methodology change in the versioned record above.
AIMC derives electricity cost and grid carbon intensity per GPU-hour from official government sources — never estimated, always cited. The full derivation, source list, and license attributions live on their dedicated pages.
We're building partnerships with GPU rental marketplaces. If you operate a marketplace or want to contribute data, let's talk.
Ai Mining Co. is a product of AIMC Technologies, LLC