GPU Rental Prices

Hourly GPU rental prices across cloud providers, matched chip for chip.

262,146price observations
40providers

Sample

2 records

Datacenter SKU, on-demand

Representative, not a real quote. Schema is reconstructed from the variables the research paper measures; there is no published schema file. Price is per GPU-hour; the panel keeps price > 0.

gpu_prices.jsonl

{
  "observed_at": "2026-05-14T08:00:00Z",
  "provider": "provider-a",
  "canonical_sku": "H100_SXM_80GB",
  "raw_instance_name": "8x H100 80GB SXM5",
  "region": "us-east",
  "price_type": "on_demand",
  "price_per_gpu_hour_usd": 2.05,
  "gpu_count": 8,
  "capacity_available": 12,
  "source": "live_poll"
}

Consumer SKU, marketplace/spot

Representative. Consumer RTX class is the long-continuous-history segment of the panel.

gpu_prices.jsonl

{
  "observed_at": "2026-05-14T08:05:00Z",
  "provider": "provider-b",
  "canonical_sku": "RTX_4090_24GB",
  "raw_instance_name": "RTX 4090 24G",
  "region": "eu-central",
  "price_type": "spot",
  "price_per_gpu_hour_usd": 0.31,
  "gpu_count": 1,
  "source": "archived_snapshot"
}

Specifications

GPU Rental Prices specifications
Formats	Parquet, JSON, CSV
Delivery	S3, REST API, Parquet
Cadence	Hourly snapshots, daily feature build
Coverage	2024 to 2026
Price observations	262,146 (six-month reference sample)
Providers	40 (AWS, Azure, GCP, CoreWeave, Crusoe, Nebius, Lambda, Vast, RunPod, io.net + 30 more)
Canonical SKUs	33 (GB200, B200, H200, H100, MI300X, A100, L40S, consumer RTX...)
Granularity	Hourly spot + on-demand, multi-region
Regions	AWS 17, Azure 57, plus marketplace geo strings
Companion tables	Utilization 1,366 rows; inference token pricing 2,397 quotes across 381 models
Provenance	Row-level source URL, collection method, raw payload pointer
Access	For internal research and model development; redistribution requires a separate agreement. Sourced from public pricing surfaces; no PII or MNPI.

What ships

Cross-Provider Price Surface
Normalized hourly spot and on-demand rates per SKU per region across 40 providers. One schema, every cloud, with the source URL retained on every row.
Capacity & Utilization
Provider capacity snapshots - total, available, and rented instances per SKU - a direct read on supply tightness.
Equity & Token Overlay
GPU-hour pricing joined to AI-infrastructure equities, options implied volatility, and DePIN compute tokens.

Fields

10 fields

Field	Description
observed_attimestamp · UTC	When the price was polled or snapshotted.Example 2026-05-14T08:00:00Z
providerstring	Cloud or marketplace provider the quote came from.Example provider-a
canonical_skustring	Normalized GPU identifier after collapsing marketplace spellings.Example H100_SXM_80GB
raw_instance_namestring · nullable	Original provider listing string before normalization.Example 8x H100 80GB SXM5
regionstring · nullable	Provider region or zone for the quote.Example us-east
price_typestring	on_demand or spot/marketplace term.Example on_demand
price_per_gpu_hour_usdfloat · USD/GPU-hour	Normalized rental price per single GPU-hour. The panel keeps price > 0.Example 2.05
gpu_countint · GPUs · nullable	GPUs in the underlying instance, used to normalize to per-GPU.Example 8
capacity_availableint · instances · nullable	Currently available, unrented capacity for the SKU.Example 12
sourcestring · nullable	live_poll or archived_snapshot (the panel backfills from archives).Example live_poll

Quality checks

2 checks

GPU-price to AI-equity signal

Whether the cross-provider price and dispersion panel predicts the AI-infrastructure equity complex.

Build price and dispersion features; train walk-forward models against forward equity returns; validate with purged k-fold cross-validation and embargo, tested against a label-shuffled null and a naive-momentum baseline.

Honest published null. Under naive evaluation the backtest looked strong; under rigorous evaluation the edge disappears - shuffle-test p-value near 1.0. Verdict: not_ready. We would rather report a null than sell false alpha.

Cross-provider price dispersion

How widely the same accelerator is priced across the market.

Take the latest price per provider per SKU for SKUs listed by at least three providers; compute the priciest-to-cheapest ratio and coefficient of variation.

Descriptive and real: order-of-magnitude dispersion - 18.3x for H100 SXM, 96.5x for A100 SXM, 100x for RTX-4090, with coefficients of variation between roughly 0.5 and 2.6.

Ground truth

What is correct	Forward realized equity returns of the AI-infrastructure complex for the predictive test; the observed cross-provider quotes themselves for the descriptive layer.
How it is established	Walk-forward modeling of price and dispersion features against forward returns, scored under purged k-fold cross-validation with embargo and benchmarked against a label-shuffled null and a naive-momentum baseline.
Agreement	Not a human-rater task. Significance is measured against the null, and the signal fails it (shuffle-test p near 1.0).

How it is built

01
Collection
Poll each provider public pricing surface, hourly where available, and backfill from archived snapshots.
02
SKU normalization
Map heterogeneous instance names to a canonical SKU list so prices are comparable across providers - the many marketplace spellings of an 80GB SXM H100 collapse to one identifier.
03
Per-hour price basis
Normalize every quote to USD per GPU-hour and filter the panel to price greater than 0.
04
Fixed-panel point-in-time reading
Read trends within a fixed group of providers. The panel composition changes over time, so a recent median up-tick can be an artifact of new, pricier providers entering rather than a real price rise; this is treated explicitly.
05
Coverage-depth handling
Consumer GPUs have the longest continuous history; datacenter SKUs are broad in the current cross-section but thin historically. The asymmetry is handled explicitly throughout.

05RequestReplies usually within one working day

Request a sample

We send real records in the format you would receive, with the field list and where each record came from. Access for evaluation is limited and covered by a mutual NDA.

ProductGPU Rental Prices