November 2026 GPU Rental Pricing Snapshot

ยท4 min read

A per-GPU hourly rate is only a starting point. If an offer requires four or eight cards, the minimum instance can cost much more. These observations show the billing details recorded on November 1.

November 1 pricing snapshot

On mobile, scroll the table horizontally to see every configuration.

GPUProviderPer GPU / hourMinimum GPUsInstance / hourRegionAvailabilitySourceCollected UTC
RTX 5090Vast.ai$0.40072$0.8015Hong Kong, HKAvailable when collectedQuote API
RTX 5090Compshareยฅ3.991Unconfirmedcn-wlcb-01Available when collectedQuote API
B300Vast.ai$8.50034$34.0014United Kingdom, GBAvailable when collectedQuote API
B300Hyperstack$7.408$59.20CANADA-1UnconfirmedQuote API
H100 SXMTensorDock$2.25UnconfirmedUnconfirmedUnconfirmedRecorded high; confirmQuote API
H100 SXMDataCrunch$4.7931$4.793UnconfirmedUnconfirmedQuote API
H200Lambda$2.29UnconfirmedUnconfirmedUnconfirmedRecorded high; confirmQuote API
H200DataCrunch$6.251$6.25UnconfirmedUnconfirmedQuote API

Rates retain their original currency; dollars and yuan are not directly comparable. Missing minimum counts, instance rates and regions are marked unconfirmed, not inferred from a per-GPU rate.

How to read this table

This limited sample spans 03:23:57โ€“12:21:51 UTC on November 1, 2026: DataCrunch H100 was collected later; other rows are from 03:23โ€“03:24. They were not necessarily available simultaneously. It is not a price trend, market minimum or a complete provider list. Availability reflects the recorded observation; unconfirmed does not mean sold out. Storage, network and taxes are not included in these compute-rate examples. Confirm current pricing, billing and inventory at provider checkout.

Source: ComputeUnion GPU quote API. Lium's B300 observation was from October 28 and is excluded from this recent snapshot. For current quotes: RTX 5090 ยท B300 ยท H100 SXM ยท H200.

Related

โ† Back to Blog