Latitude.shIn stock
On-Demand from $1.68
From
$1.68 / GPU / hr
On-DemandVisit website →
Rental price index
NVIDIA H100 SXM
🏢The reference datacenter GPU. Underlies the CME H100 rental index.
Prices collected 2026-09-12 · 23 providers, 62 configurations
Compare vs other GPUs →On-demandOn-demand (archive, Lambda only)Spot
Index median price per GPU per hour, by billing type. Nothing is interpolated.
At a glance
Median on-demand
$3.49/GPU/hr
Cheapest on-demand
$1.68Latitude.sh
Latitude.shFloor, any billing
$1.68incl. spot
Per month at 720h
$2,513
NVIDIA H100 SXM pricing by provider
One card per provider, cheapest first. A configuration is one priced unit — an instance type, a SKU, a marketplace offer — and the per-GPU price is that unit's price divided by its GPU count. Open a card to see every configuration with its vCPU, RAM and region where the provider publishes them.
23 providers · 57 configurations
HyperAIIn stock
CNChina1 config1xcollected 2026-09-12
On-Demand from $1.80
From
$1.80 / GPU / hr
On-DemandVisit website →SpheronIn stock
2 configs1xcollected 2026-09-12
On-Demand from $2.01
From
$2.01 / GPU / hr
On-DemandVisit website →
Vast.aiIn stock
USUSA8 configs1x–2xcollected 2026-09-12
On-Demand from $2.27
From
$2.27 / GPU / hr
On-Demand · The Netherlands, NLVisit website →
OblivusIn stock
GBUK6 configs1x–8xcollected 2026-09-12
On-Demand from $2.49
From
$2.49 / GPU / hr
On-DemandVisit website →
LyceumIn stock
DEGermany3 configs1xcollected 2026-09-12
On-Demand from $2.49
From
$2.49 / GPU / hr
On-DemandVisit website →
KoyebIn stock
FRFrance4 configs1x–8xcollected 2026-09-12
On-Demand from $2.50
From
$2.50 / GPU / hr
On-DemandVisit website →
RunPod Community CloudIn stock
USUSA1 config1xcollected 2026-09-12
On-Demand from $2.69
From
$2.69 / GPU / hr
On-DemandVisit website →Massed ComputeIn stock
USUSA2 configs1xcollected 2026-09-12
On-Demand from $2.73
From
$2.73 / GPU / hr
On-DemandVisit website →
RuncrateIn stock
1 config1xcollected 2026-09-12
On-Demand from $2.75
From
$2.75 / GPU / hr
On-DemandVisit website →
CivoIn stock
GBUK2 configs1x–8xcollected 2026-09-12
On-Demand from $2.99
From
$2.99 / GPU / hr
On-DemandVisit website →
GPU.aiIn stock
AEUAE2 configs1xcollected 2026-09-12
On-Demand from $3.06
From
$3.06 / GPU / hr
On-DemandVisit website →
Paperspace (DigitalOcean)In stock
USUSA2 configs1xcollected 2026-09-12
On-Demand from $3.09
From
$3.09 / GPU / hr
On-DemandVisit website →
DataCrunch (Verda)In stock
FIFinland2 configs1xcollected 2026-09-12
On-Demand from $3.25
From
$3.25 / GPU / hr
On-DemandVisit website →
RunPod Secure CloudIn stock
USUSA1 config1xcollected 2026-09-12
On-Demand from $3.49
From
$3.49 / GPU / hr
On-DemandVisit website →
BeamIn stock
USUSA1 config1xcollected 2026-09-12
On-Demand from $3.50
From
$3.50 / GPU / hr
On-DemandVisit website →
Cyfuture AIIn stock
INIndia1 config8xcollected 2026-09-12
On-Demand from $3.56
From
$3.56 / GPU / hr
On-DemandVisit website →
LambdaIn stock
USUSA4 configs1x–8xcollected 2026-09-12
On-Demand from $3.99
From
$3.99 / GPU / hr
On-DemandVisit website →
fal.aiIn stock
USUSA1 config1xcollected 2026-09-12
On-Demand from $4.50
From
$4.50 / GPU / hr
On-DemandVisit website →
ReplicateIn stock
USUSA4 configs1x–8xcollected 2026-09-12
On-Demand from $5.49
From
$5.49 / GPU / hr
On-DemandVisit website →
CoreWeaveIn stock
USUSA1 config8xcollected 2026-09-12
On-Demand from $6.16
From
$6.16 / GPU / hr
On-DemandVisit website →
Amazon Web ServicesIn stock
USUSA2 configs1x–8xcollected 2026-09-12
On-Demand from $6.88
From
$6.88 / GPU / hr
On-Demand · us-east-1Visit website →
Microsoft AzureIn stock
USUSA5 configs8xcollected 2026-09-12
On-Demand from $11.06Spot from $2.04
From
$11.06 / GPU / hr
On-Demand · eastusVisit website →What fits on a NVIDIA H100 SXM
Weights + KV cache at 8K context + runtime overhead, against 72 GB usable of the 80 GB on the card. Capacity follows total parameters, even for mixture-of-experts models — any expert may be needed next. Where one card is not enough, the table says how many are, and prices the set at today's median.
| Model | Params | 4-bit | 8-bit | 16-bit |
|---|---|---|---|---|
| 2.8T104B active | 22× GPU$76.78/hr | 42× GPU$146.58/hr | 78× GPU$272.22/hr | |
| 552B16B active | 5× GPU$17.45/hr | 9× GPU$31.41/hr | 16× GPU$55.84/hr | |
 Llama 4 Maverick | 400B17B active | 4× GPU$13.96/hr | 6× GPU$20.94/hr | 12× GPU$41.88/hr |
 gpt-oss-120b | 117B5.1B active | ✓ 67.9 GB~934 tok/s | 2× GPU$6.98/hr | 4× GPU$13.96/hr |
| Llama 3.1 70B | 70B | ✓ 44.5 GB~68 tok/s | 2× GPU$6.98/hr | 3× GPU$10.47/hr |
 Gemma 4 31B | 31B | ✓ 21.2 GB~154 tok/s | ✓ 36.7 GB~81 tok/s | ✓ 65.7 GB~43 tok/s |
 Qwen3 Coder 30B-A3B | 30.5B3.3B active | ✓ 19 GB~1444 tok/s | ✓ 34.3 GB~764 tok/s | ✓ 62.9 GB~406 tok/s |
| Qwen3.8 27B | 27B | ✓ 18.7 GB~176 tok/s | ✓ 32.2 GB~93 tok/s | ✓ 57.6 GB~50 tok/s |
| gpt-oss-20b | 21B3.6B active | ✓ 13.7 GB~1323 tok/s | ✓ 24.2 GB~701 tok/s | ✓ 43.9 GB~372 tok/s |
| Llama 3.1 8B | 8B | ✓ 6.9 GB~596 tok/s | ✓ 10.9 GB~315 tok/s | ✓ 18.4 GB~168 tok/s |
tok/s is decode throughput from memory bandwidth, derated 20%, single stream — an estimate, not a benchmark. Set cost uses today's median × GPUs required.
Specifications
Memory80 GB HBM3
Memory bandwidth3,350 GB/s
ArchitectureHopper
VendorNVIDIA
ClassDatacenter
Released2022-09
Bandwidth is listed because it governs generation speed: each token streams the active weights once, so tokens per second is bandwidth divided by bytes read. TFLOPs decide prompt processing and training, not how fast text appears.
Alternatives
NVIDIA A100 80GB80 GB · $1.79/hr median · 19 providersNVIDIA H100 PCIe80 GB · $2.51/hr median · 7 providersNVIDIA H100 NVL94 GB · $2.95/hr median · 6 providers
Questions
How much does it cost to rent a NVIDIA H100 SXM?
As of 2026-09-12, the median on-demand price across 23 providers is $3.49 per GPU-hour, about $2,513 a month at 720 hours. The cheapest we saw was $1.68 at Latitude.sh.
Why do prices for the same GPU vary so much?
Hyperscalers bundle CPU, RAM and network into an instance price and charge for the ecosystem around it. Neoclouds sell the GPU more directly. Marketplaces are independent hosts competing on price, with the trade-offs of shared, third-party hardware. The per-GPU figure is the instance price divided by GPU count, which is the only way to compare an 8-GPU node with a single-card listing.
Is the median the price I will pay?
No — it is the middle of the market. Each provider's own median is taken first so a provider with many instance sizes counts once, then the median across providers. Spot and reserved rates are excluded from it; they set the floor shown separately. Use it to judge whether a quote is high or low, then verify with the provider.
Which models can a NVIDIA H100 SXM run?
With 80 GB per GPU, the table above shows which reference models fit on one card at 4, 8 and 16-bit, using the same memory arithmetic as the rest of Runyard: weights plus KV cache plus runtime overhead, with capacity following total parameters even for mixture-of-experts models. Where a model needs more than one GPU it says how many and what that set costs at today's median.
Own the hardware instead? Work out what your card holds with the same arithmetic.
Check what fits →