Below $1.50 an hour, the cheapest H100s are spot bids and thin listings — and hyperscaler egress quietly adds up to $1.20 an hour

Asked:

“What is the real effective price per hour for an NVIDIA H100 across GPU cloud providers? Include the sticker price, spot or committed discounts, egress fees, and current availability for each provider.”

This comparison covers 24 provider/offering rows observed in 2026, current through 2026-09-07, all prices in USD per H100 GPU-hour. It spans marketplaces, GPU-specialist clouds, and the four hyperscalers; each price, discount, egress fee, and availability statement links its own source. Several rates are triangulated from secondary comparison pages, and marketplace rows change with inventory.

guaranteed / on-demand sticker spot / preemptible / marketplace (interruptible) committed / reserved term + 10 GB paid egress in the hour sticker price marker
IO.NET’s $1.20 spot bid is the lowest supported compute rate of all 24 rows, but its $0.02/GB egress lifts the 10 GB scenario to $1.40 — and it is one of the thinnest-sourced rows here (3 independent hosts). misar.blog
Hyperbolic ($1.29) and ThunderCompute ($1.38) are the cheapest straightforward sticker rates with $0/GB egress — no interruption risk and no marketplace variability. gpufinder.dev, intuitionlabs.ai
Hyperscaler spot looks competitive — Azure and GCP at $2.25, AWS p5 at $3.00 — until egress is counted: 10 GB in the hour pushes them to $3.12, $3.45, and $3.90, above nearly every specialist cloud. intuitionlabs.ai
AWS Capacity Blocks at $4.72 is an advance-reservation product, not another on-demand provider — you book a window ahead of time; OCI’s best rate needs a 1-year commitment and still lands at $6.59 with 10 GB egress. aws.amazon.com, thundercompute.com

All 24 offerings

ProviderConfigurationSticker $/hDiscount $/hTermEgress $/GBFree egressEffective $/h+10 GB $/hAvailabilitySource

Method: 24 provider/offering rows for NVIDIA H100 GPU-hour pricing, observed 2026 (latest 2026-09-07), all values USD per GPU-hour. “Effective” is the lower currently supported spot/committed rate when no higher than sticker, else sticker; “+10 GB” adds 10× the first paid-tier public-internet egress fee as a sensitivity scenario, not a workload assumption. Free monthly allowances are shown but not amortized. Excluded: CPU/RAM/storage, taxes, inter-zone traffic, minimum billing, marketplace host fees, interruption/restart costs. H100 variants and bundled host resources differ; do not read as benchmark-equivalent. Several rates come from secondary comparison pages; rows backed by 5 or fewer independent hosts, or marketplace-variable, are flagged. AWS Capacity Blocks is an advance-reservation product listed separately from AWS EC2 p5.

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.

Keenable SELECTAsk your own question
Made with Keenable SELECT