io.netIn stock
On-Demand from $1.59
From
$1.59 / GPU / hr
On-DemandVisit website →
Rental price index
NVIDIA H100 PCIe
🏢The PCIe H100. Same memory as the SXM, roughly 60% of its bandwidth.
Prices collected 2026-10-11 · 8 providers, 12 configurations
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
$2.29/GPU/hr
Cheapest on-demand
$1.59io.net
io.netFloor, any billing
$1.59incl. spot
Per month at 720h
$1,646
NVIDIA H100 PCIe 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.
8 providers · 11 configurations
BeamIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $1.83
From
$1.83 / GPU / hr
On-DemandVisit website →
GPU.aiIn stock
AEUAE2 configs1xcollected 2026-10-11
On-Demand from $1.98
From
$1.98 / GPU / hr
On-DemandVisit website →
RunPod Community CloudIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $1.99
From
$1.99 / GPU / hr
On-DemandVisit website →
Vast.aiIn stock
USUSA3 configs1x–2xcollected 2026-10-11
On-Demand from $2.29
From
$2.29 / GPU / hr
On-Demand · South Africa, ZAVisit website →
RunPod Secure CloudIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $2.89
From
$2.89 / GPU / hr
On-DemandVisit website →
Cyfuture AIIn stock
INIndia1 config1xcollected 2026-10-11
On-Demand from $2.99
From
$2.99 / GPU / hr
On-DemandVisit website →
LambdaIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $3.29
From
$3.29 / GPU / hr
On-DemandVisit website →Recent price changes
Every time a provider's median for the NVIDIA H100 PCIe moved between two consecutive snapshots. A listing appearing or vanishing is not a price change and is not shown.
| When | Provider | From | To | Change |
|---|---|---|---|---|
| 2026-10-11 09:14 | GPU.ai | $2.13 | $2.29 | +7.5% |
| 2026-10-10 08:53 | Vast.ai | $2.30 | $2.27 | -1.6% |
| 2026-10-09 09:31 | Vast.ai | $2.66 | $2.30 | -13.2% |
| 2026-10-08 09:24 | Vast.ai | $2.31 | $2.66 | +15.1% |
| 2026-10-07 09:10 | GPU.ai | $2.48 | $1.98 | -20.2% |
| 2026-10-07 09:10 | Vast.ai | $2.00 | $2.31 | +15.3% |
| 2026-10-06 09:19 | Vast.ai | $2.59 | $2.00 | -22.7% |
| 2026-10-06 09:19 | GPU.ai | $2.29 | $2.48 | +8.5% |
| 2026-10-05 09:33 | GPU.ai | $1.98 | $2.29 | +15.4% |
| 2026-10-03 08:31 | Vast.ai | $2.33 | $2.59 | +10.9% |
Spot, reserved and where it is cheapest
Spot median
$2.0013% off · 1 providers
Reserved median
—mostly by quote
USA providers
$2.14median of 6
UAE providers
$2.29median of 1
By provider country: USA $2.14 (6) · UAE $2.29 (1) · India $2.99 (1). Full breakdown →
What fits on a NVIDIA H100 PCIe
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$50.31/hr | 42× GPU$96.04/hr | 78× GPU$178.36/hr | |
| 552B16B active | 5× GPU$11.43/hr | 9× GPU$20.58/hr | 16× GPU$36.59/hr | |
 Llama 4 Maverick | 400B17B active | 4× GPU$9.15/hr | 6× GPU$13.72/hr | 12× GPU$27.44/hr |
 gpt-oss-120b | 117B5.1B active | ✓ 67.9 GB~558 tok/s · $1.14/M tok | 2× GPU$4.57/hr | 4× GPU$9.15/hr |
| Llama 3.1 70B | 70B | ✓ 44.5 GB~41 tok/s · $15.49/M tok | 2× GPU$4.57/hr | 3× GPU$6.86/hr |
 Gemma 4 31B | 31B | ✓ 21.2 GB~92 tok/s · $6.9/M tok | ✓ 36.7 GB~49 tok/s · $12.96/M tok | ✓ 65.7 GB~26 tok/s · $24.43/M tok |
 Qwen3 Coder 30B-A3B | 30.5B3.3B active | ✓ 19 GB~862 tok/s · $0.74/M tok | ✓ 34.3 GB~456 tok/s · $1.39/M tok | ✓ 62.9 GB~242 tok/s · $2.62/M tok |
| Qwen3.8 27B | 27B | ✓ 18.7 GB~105 tok/s · $6.05/M tok | ✓ 32.2 GB~56 tok/s · $11.34/M tok | ✓ 57.6 GB~30 tok/s · $21.17/M tok |
| gpt-oss-20b | 21B3.6B active | ✓ 13.7 GB~790 tok/s · $0.8/M tok | ✓ 24.2 GB~418 tok/s · $1.52/M tok | ✓ 43.9 GB~222 tok/s · $2.86/M tok |
| Llama 3.1 8B | 8B | ✓ 6.9 GB~356 tok/s · $1.78/M tok | ✓ 10.9 GB~188 tok/s · $3.38/M tok | ✓ 18.4 GB~100 tok/s · $6.35/M tok |
tok/s is decode throughput from memory bandwidth, derated 20%, single stream — an estimate, not a benchmark. $/M tok is the hourly median divided by that rate: the one-user floor for a million generated tokens. Set cost uses today's median × GPUs required.
Buy or rent a H100 PCIe?
Approx. list price
$24,000not tracked
Rent, per month
$1,646at median, 720 h
Break-even
10,495 h≈ 14.6 months
Rent for fewer than about 14.6 months of continuous use and renting wins on the card alone; past that, owning does — before the server, power, cooling and the person who keeps it running, which for a datacenter card are not small. The list price is a typical quoted figure and is not tracked by the index.
Specifications
Memory80 GB HBM2e
Memory bandwidth2,000 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 and comparisons
NVIDIA H100 SXM80 GB · $3.50/hr median · 43 providersNVIDIA A100 80GB80 GB · $1.79/hr median · 35 providersNVIDIA H100 NVL94 GB · $2.80/hr median · 6 providers
Questions
How much does it cost to rent a NVIDIA H100 PCIe?
As of 2026-10-11, the median on-demand price across 8 providers is $2.29 per GPU-hour, about $1,646 a month at 720 hours. The cheapest we saw was $1.59 at io.net.
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 PCIe 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 →