H100 · spot and reserved
Read 12 November 2026
H100 spot and reserved prices
- Spot lowest
- $1.42/GPU-hr
- Microsoft Azure: 29% below the lowest on-demand price
- Reserved lowest
- $1.75/GPU-hr
- Hyperstack: 12% below the lowest on-demand price
- On-demand lowest
- $1.99/GPU-hr
- Voltage Park
- Rows
- 19
- 12 spot, 7 reserved
Basis: the lowest row of each type where the provider sets the price, against the lowest solid on-demand row. Marketplace and from figures appear in the tables with a chip but not in these readings. USD per GPU-hour, before tax.
What the terms mean
definitions, not claims about any provider- On-demand
- Rent by the hour or smaller unit with no term. This is the basis of the rate card and of every statistic on it.
- Spot
- Capacity sold at a price that can change, where the provider may interrupt or reclaim it. Some providers call it preemptible or interruptible.
- Reserved
- Capacity committed and paid for over a term. The rate and the term length are whatever the provider states; the row description names the term when the page does, and the term is not stated otherwise.
- Difference
- Percent change from the same provider’s lowest solid on-demand H100 price to the row: (row minus on-demand) divided by on-demand. A minus sign means lower than on-demand.
Each provider's three prices
8 providersA row says none where the provider has no row of that type or its figure is a marketplace or from figure.
Spot rows
12 rows| Provider | Configuration | Billing | $/GPU-hr | Its on-demand low | Difference | Read | Source |
|---|---|---|---|---|---|---|---|
| Nebiusmarketplace | 8× HGXNVIDIA HGX H100 (preemptible) | not stated | $0.79 | $4.50 | -82% | 2026-11-12 | page |
| Microsoft Azure | 1× NVLStandard_NC40ads_H100_v5 (Spot) | not stated | $1.42 | $6.98 | -80% | 2026-11-12 | page |
| Microsoft Azure | 2× NVLStandard_NC80adis_H100_v5 (Spot) | not stated | $1.42 | $6.98 | -80% | 2026-11-12 | page |
| Together AI | 8× HGXNVIDIA HGX H100 cluster preemptible | not stated | $1.99 | $3.99 | -50% | 2026-11-12 | page |
| Hyperstack | 1× PCIeNVIDIA H100 PCIe Spot VM | per minute | $2.00 | $2.50 | -20% | 2026-11-12 | page |
| Microsoft Azure | 8×Standard_ND96isr_H100_v5 (Spot) | not stated | $2.27 | $6.98 | -67% | 2026-11-12 | page |
| CoreWeave | 8× HGXNVIDIA HGX H100 | not stated | $2.44 | $6.16 | -60% | 2026-11-12 | page |
| CoreWeave | 8× HGXNVIDIA HGX H100 | not stated | $2.46 | $6.16 | -60% | 2026-11-12 | page |
| Verda | 1× SXM1x H100 SXM5 80GB (30 vCPU, 120 GB RAM) spot | not stated | $2.54 | $5.09 | -50% | 2026-11-12 | page |
| Verda | 1× SXM1x H100 SXM5 80GB (32 vCPU, 170 GB RAM) spot | not stated | $2.54 | $5.09 | -50% | 2026-11-12 | page |
| Google Cloud | 8×a3-megagpu-8g (208 vCPU, 1,872GB RAM) | not stated | $6.28 | $11.06 | -43% | 2026-11-12 | page |
| Google Cloud | 8×a3-highgpu-8g (208 vCPU, 1,871GB RAM) | not stated | $6.59 | $11.06 | -40% | 2026-11-12 | page |
12 of 12 spot rows that can be compared are below the same provider's lowest on-demand price. The difference runs from -82% (Nebius) to -20% (Hyperstack). 1 row (Nebius) is a marketplace or "from" figure and not counted in the lowest spot price.
Reserved rows
7 rows| Provider | Configuration | Billing | $/GPU-hr | Its on-demand low | Difference | Read | Source |
|---|---|---|---|---|---|---|---|
| Hyperstack | 1× PCIeNVIDIA H100 reservation | per minute | $1.75 | $2.50 | -30% | 2026-11-12 | page |
| Hyperstack | 1×NVIDIA H100 NVLink (NVLink) reservation | per minute | $1.82 | $2.50 | -27% | 2026-11-12 | page |
| Vultr | 8× SXMBare Metal 8 x NVIDIA H100 80 GB SXM | not stated | $2.30 | $2.30 | +0% | 2026-11-12 | page |
| Hyperstack | 1× SXMNVIDIA H100 SXM reservation | per minute | $2.72 | $2.50 | +9% | 2026-11-12 | page |
| DigitalOcean (Paperspace) | 1× HGXNVIDIA HGX H100 12-month reserved | not stated | $3.26 | $4.41 | -26% | 2026-11-12 | page |
| Together AI | 8× HGXNVIDIA HGX H100 cluster reserved 31-90 days | not stated | $3.45 | $3.99 | -14% | 2026-11-12 | page |
| Together AI | 8× HGXNVIDIA HGX H100 cluster reserved 3-30 days | not stated | $3.69 | $3.99 | -8% | 2026-11-12 | page |
5 of 7 reserved rows that can be compared are below the same provider's lowest on-demand price. The difference runs from -30% (Hyperstack) to +9% (Hyperstack). Where a row names no term in its description, the term is not stated on the page we read.
Spot and reserved: common questions
Is spot cheaper than on-demand?
12 of 12 spot rows are below the same provider's lowest on-demand price, by as much as -82% at Nebius. Spot capacity can be interrupted, so the rate is not like for like with on-demand.
What is the lowest spot and the lowest reserved H100 price?
Spot: $1.42 at Microsoft Azure, 29% below the lowest on-demand price of $1.99. Reserved: $1.75 at Hyperstack, 12% below the lowest on-demand price.
How much lower is reserved than on-demand at the same provider?
Comparable reserved rows differ from the same provider's lowest on-demand price by -30% (Hyperstack) to +9% (Hyperstack). Read the term on the provider's page before relying on a row.
Why are some rows marked marketplace?
A marketplace or "from" row is a figure set by hosts, or the lowest of a range, not a fixed list price. 1 commitment rows carry the chip. They are shown but left out of the lowest-price readings.
Does Runcrate list spot or reserved H100 capacity?
Its live feed shows 4 prepaid terms for the H100, listed above and marked as ours. There is no spot row.
Disclosure
H100 Rate Card is published by the Runcrate team, who also run Runcrate, an AI cloud. Runcrate is listed in the table for the H100, marked "ours, live". It is sorted by price like every other row, with no pinned position. Other providers' prices are read from their own public pricing pages. We have no paid placement.