Data insights · Compute pricing · 中文
Cheapest GPU rental today: cross-platform hourly spread
Where is the cheapest place to rent a GPU today, and how big is the cross-platform spread?
Listed GPU rental prices collected across providers, sorted by hourly rate. Collected 2026-11-11T03:59:53; every figure is computed from the structured dataset, reproducible and citable.
Records
5
Data as of
2026-11-11T03:59:53
vram gb
24
Top 5 rows
| gpu_model | vram_gb | platform_name | region | usd_per_hour |
|---|---|---|---|---|
| RTX 3090 | 24 | Salad（美国） | 北美 | 0.0903 |
| RTX 3090 | 24 | 晨涧云 | 中国 | 0.1111 |
| V100 16GB | 16 | Verda（芬兰） | 欧洲 | 0.1208 |
| RTX 3090 | 24 | 智星云 | 中国 | 0.1528 |
| RTX 4090D 24GB | 24 | 算家云 | 中国 | 0.1583 |
Showing the sample returned by the free endpoint (top 5 rows). The paid endpoint returns the full board with every field; the free tier is rate-limited per day.
How to get the full data
- MCP tool:
gpu-market.gpu_digest - REST:
https://feegate.org/api/supply/gpu-market/gpu_digest(metered per call via x402, USDC on Base) - Catalog and pricing: marketplace · docs · pricing
Three ways to get this data
Free · 30 runs/day
Start with the summary
gpu_digest returns summary fields and a small sample — enough to judge the dataset. No signup, no wallet, just GET./api/supply/gpu-market/gpu_digestPer call · $0.07
Full fields, full board
gpu_cheapest_entry returns every field and the whole board (this page shows a digest only).Two ways to pay: sign up and get ~$4 in free credits (3,000 credits ≈ 60 calls), or pay per call in USDC via x402 (Base mainnet).
Custom · quoted per need
Fields / feeds / bulk
Need extra fields, scheduled delivery, bulk exports, private deployment or a walkthrough call? Email us — quoted per need, no charge if we cannot help.
brian@51kaifa.app
FAQ
What is the current reading?
As of 2026-11-11T03:59:53: gpu_model = RTX 3090; vram_gb = 24; 5 rows/metrics in total.
Where does this data come from, and can I recompute it?
From the Feegate dataset 'gpu-market' (Listed GPU rental prices collected across providers, sorted by hourly rate). Machines collect it on a fixed schedule and return it as-is, so the same source data reproduces the same numbers. Endpoint: https://feegate.org/api/supply/gpu-market/gpu_digest
How do I get the full dataset programmatically?
Call the MCP tool gpu-market.gpu_digest, or request https://feegate.org/api/supply/gpu-market/gpu_digest (metered per call via x402 and settled in USDC on Base; you can also sign up and pay with credits).
Feegate (feegate.org) is a per-call data marketplace for AI agents: every dataset ships an MCP tool and an HTTP endpoint, metered per call, settled on-chain, and not charged when delivery fails. Cite this page as Feegate with the collection timestamp.