H200 · 141 GB · Hopper

Every H200 price, in one ledger, with its source.

As of 12 November 2026, 15 providers list an on-demand H200 between $3.49 and $10.60 per GPU-hour. The median of each provider's lowest price is $5.40. Each figure is read from the provider's own page.

Lowest listed
$3.49

Civo, per GPU-hour

Median
$5.40

of provider lows

Highest
$10.60

Google Cloud

ProviderPriceStep
  1. Civo$3.49lowest
  2. Massed Compute$3.62+$0.13
  3. Hyperstack$3.99+$0.37
  4. Crusoe$4.29+$0.30
  5. DigitalOcean (Paperspace)$4.47+$0.18
  6. Runcrateours, live$4.67+$0.20
  7. RunPod$5.29+$0.62
  8. Median of 15 providers$5.40
  9. Nebius$5.40+$0.11
  10. Together AI$5.99+$0.59
  11. OVHcloud$6.20+$0.21
  12. CoreWeave$6.30+$0.11
  13. Verda$6.50+$0.20
  14. Amazon Web Services$7.91+$1.41
  15. Oracle Cloud$10.00+$2.09
  16. Google Cloud$10.60+$0.60
Each rung is one provider's lowest solid on-demand H200 price per GPU-hour, on a true scale from $3.49 to $10.60. The step is the gap to the rung below. Marketplace and announced prices are left out.

What an 8-GPU server costs

All 9 providers with 8-GPU servers

The lowest solid 8-GPU on-demand configuration in the data, multiplied out. The highest is Google Cloud at $84.81 per server-hour, 3.0 times the lowest.

Per GPU-hour
$3.49

Civo, 8× SXM

Per server-hour
$27.92

$3.49 × 8 GPUs

Per day
$670.08

$27.92 × 24 hours

Per 720-hour month
$20,102.40

$27.92 × 720 hours

The ledger

Every current row, grouped by size

Rows are the on-demand configurations in effect today, grouped by GPUs per server. Per server-hour is the GPU price times the GPUs in the row.

Marketplace figures carry a chip and are left out of every statistic.

29 rows in 4 sizes
Current on-demand H200 prices grouped by GPUs per server
ConfigurationBillingRegionSource
Single GPU14 rowslowest solid: $3.49 per server-hour at Civo
Vast.aimarketplace figure1× · Marketplace lowest listed (H200)not stated$2.63$2.63not stated2026-11-12page
Civo1× SXM · Small 1 x NVIDIA H200 - 141GB HBM3eper hour$3.49$3.49not stated2026-11-12page
RunPodmarketplace figure1× SXM · Community Cloud H200per second$3.59$3.59not stated2026-11-12page
Vast.aimarketplace figure1× NVL · Marketplace lowest listed (H200 NVL)not stated$3.60$3.60not stated2026-11-12page
Massed Compute1× NVL · H200 NVL (141GB)not stated$3.62$3.62not stated2026-11-12page
Hyperstack1× SXM · NVIDIA H200 SXMper minute$3.99$3.99not stated2026-11-12page
Vast.aimarketplace figure1× NVL · Marketplace median (H200 NVL)not stated$4.39$4.39not stated2026-11-12page
DigitalOcean (Paperspace)1× HGX · NVIDIA HGX H200 on-demandper second$4.47$4.47not stated2026-11-12page
Runcrateours, live1× NVLper minute$4.67$4.67Worcester, United Kingdom2026-11-12page
Runcrateours, live1×per second$5.13$5.134 regions2026-11-12page
Vast.aimarketplace figure1× · Marketplace median (H200)not stated$5.20$5.20not stated2026-11-12page
RunPod1× SXM · Secure Cloud H200per second$5.29$5.29not stated2026-11-12page
Verda1× SXM · 1x H200 SXM5 141GB (44 vCPU, 170 GB RAM)not stated$6.50$6.50not stated2026-11-12page
Oracle Cloud1× · OCI - Compute - GPU - H200 (B110519)not stated$10.00$10.00not stated2026-11-12page
2 GPUs per server1 rowlowest solid: $7.24 per server-hour at Massed Compute
Massed Compute2× NVL · H200 NVL (141GB)not stated$3.62$7.24not stated2026-11-12page
4 GPUs per server2 rowslowest solid: $14.48 per server-hour at Massed Compute
Massed Compute4× NVL · H200 NVL (141GB)not stated$3.62$14.48not stated2026-11-12page
Massed Compute4× NVL · H200 NVL (141GB) NVLinknot stated$3.82$15.28not stated2026-11-12page
8 GPUs per server12 rowslowest solid: $27.92 per server-hour at Civo
Civo8× SXM · Extra Large 8 x NVIDIA H200 - 141GB HBM3eper hour$3.49$27.92not stated2026-11-12page
Massed Compute8× NVL · H200 NVL (141GB)not stated$3.62$28.96not stated2026-11-12page
Massed Compute8× NVL · H200 NVL (141GB) NVLinknot stated$3.82$30.56not stated2026-11-12page
Crusoe8× HGX · NVIDIA H200 HGXper hour$4.29$34.32not stated2026-11-12page
Nebius8× HGX · NVIDIA HGX H200not stated$5.40$43.20not stated2026-11-12page
Together AI8× HGX · NVIDIA HGX H200 cluster on-demandnot stated$5.99$47.92not stated2026-11-12page
OVHcloud8× · h200-1920per hour$6.20$49.56not stated2026-11-12page
OVHcloud8× · h200-1920-ephper hour$6.20$49.56not stated2026-11-12page
CoreWeave8× HGX · NVIDIA HGX H200not stated$6.30$50.44North America2026-11-12page
CoreWeave8× HGX · NVIDIA HGX H200not stated$6.30$50.44Europe2026-11-12page
Amazon Web Services8× · p5en.48xlargeper second$7.91$63.30us-east-12026-11-12page
Google Cloud8× · a3-ultragpu-8g (224 vCPU, 2,952GB RAM)not stated$10.60$84.81not stated2026-11-12page

Announced, not in effect

  • Crusoe: 8× HGX at $5.40 per GPU-hour ($43.20 per server-hour) from 2026-12-01. source

Prices with a later start date are shown here and kept out of the statistics until they take effect.

Terms

On-demand, spot and reserved

Three different promises. The figures below are each type's lowest solid listing, from different providers and sizes, so read them as a guide, not a like-for-like trade.

On-demand

$3.49

Civo, 1× SXM

No commitment, billed as used.

Spot or preemptible

$2.58

CoreWeave, 8× HGX (-26% against the lowest on-demand)

Can be reclaimed by the provider.

Reserved

$2.79

Hyperstack, 1× SXM (-20% against the lowest on-demand)

A contract term. Term length is shown where stated.

Per-provider comparison of on-demand, spot and reserved

Budget

What a month of budget buys

At the lowest solid price, $3.49 per GPU-hour, one GPU running a 720-hour month costs $2,512.80.

$5,000 a month

1 GPU

for a full 720-hour month: $5,000 ÷ ($3.49 × 720) = 1.99, rounded down

$20,000 a month

7 GPUs

for a full 720-hour month: $20,000 ÷ ($3.49 × 720) = 7.96, rounded down

$50,000 a month

19 GPUs

for a full 720-hour month: $50,000 ÷ ($3.49 × 720) = 19.90, rounded down

The same budgets at the median and the highest price, and per provider

Questions

H200 prices, answered from the data

What is the lowest H200 price per hour?
$3.49 per GPU-hour at Civo (1× SXM), read 2026-11-12. Across the 15 providers that set their own price, the median of each provider's lowest price is $5.40 and the highest is $10.60 at Google Cloud, 3.0 times the lowest.
How much does an 8-GPU H200 server cost per hour?
At the lowest solid 8-GPU configuration, Civo, it is $3.49 per GPU times 8 = $27.92 per hour, $670.08 per day and $20,102.40 per 720-hour month. The highest of the 9 providers with an 8-GPU row is Google Cloud at $84.81 per hour.
Why is the spread between providers so wide?
The rungs run from $3.49 to $10.60. Sizes differ (single GPUs and 8-GPU servers sit in the same ladder), billing differs, and for Amazon Web Services ($7.91) and Google Cloud ($10.60) the recorded note says the instance price includes vCPUs and memory, so the per-GPU figure is not a GPU-only price. The table shows size, billing and region next to every price so you can compare like with like.
Where does Runcrate sit?
Runcrate's live H200 price is $4.67 per GPU-hour, rank 6 of 15 by price and -14% against the median. It is placed by the same rule as every other row and marked "ours, live".
Are marketplace prices included?
5 current rows are marketplace figures (Vast.ai, RunPod). The host sets those, not the provider, so they carry a chip in the table and are left out of the ladder, the median and every ranking.
How current are these prices?
The stored prices were last read on 12 November 2026, and each row shows its own read date and source page. Runcrate's figure comes live from runcrate.ai. This is one snapshot: the ledger shows no history.

Read on

More ways into the same data

H200 against another GPU

One page per provider

Disclosure

H200 Price Ledger is published by the Runcrate team, who also run Runcrate, an AI cloud. Runcrate is listed in the table for H200, 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.