H100 PCIe from $0.45 per GPU-hour
The PCIe H100 trades some bandwidth and interconnect for a much lower price. For single-GPU inference or fine-tuning that never leaves one card, the difference against the SXM board is small and the saving is not.
Billed per started minute. 640 GB of GPU memory on a 8× node. Persistent storage $0.08 per GB-month. No egress fees.
- GPU memory
- 80GBHBM2e · MIG capable
- Bandwidth
- 2,000GB/sSets token throughput on inference
- Node sizes
- 1× · 2× · 4× · 8×Same price per GPU at every size
- Reclaim rate
- <5%Spot instances reclaimed over 30 days
What you get with each H100 PCIe node.
| Node | GPU memory | vCPU | RAM | Local NVMe | Spot | On-demand | Action |
|---|---|---|---|---|---|---|---|
| 1× H100 PCIe | 80 GB | 26 | 240 GB | 2 TB | $0.45/h | $0.69/h | Configure |
| 2× H100 PCIe | 160 GB | 52 | 480 GB | 4 TB | $0.90/h | $1.38/h | Configure |
| 4× H100 PCIe | 320 GB | 104 | 960 GB | 8 TB | $1.80/h | $2.76/h | Configure |
| 8× H100 PCIe | 640 GB | 208 | 1,920 GB | 16 TB | $3.60/h | $5.52/h | Configure |
GPU-to-GPU link: PCIe Gen5 x16; NVLink bridge (3 bridges) 600 GB/s to one adjacent card. Local NVMe is scratch space wiped when the instance ends; keep anything you need on a persistent disk.
96 H100 PCIe GPUs free right now.
Straight from the capacity pool the console books against. Prices are identical in every region — pick the one closest to your data.
A region with zero free GPUs still accepts reserved capacity requests — we hold hardware for a term rather than sell what is already taken.
What fits in 80 GB — and what it costs per hour.
Weights need about 2 GB per billion parameters at FP16 / BF16, 1 GB per billion parameters at FP8, 0.5 GB per billion parameters at INT4. We keep 20% of the memory free for activations, the KV cache and the CUDA context, then take the smallest node that still fits.
| Open model | Parameters | FP16 / BF16node · spot price | FP8node · spot price | INT4node · spot price |
|---|---|---|---|---|
| Mistral 7BAssistants, RAG, classification | 7.2B | 1× $0.45/h | 1× $0.45/h | 1× $0.45/h |
| Llama 3.1 8BAssistants, agents, fine-tuning | 8B | 1× $0.45/h | 1× $0.45/h | 1× $0.45/h |
| Gemma 2 27BHigher-quality assistants | 27B | 1× $0.45/h | 1× $0.45/h | 1× $0.45/h |
| Qwen2.5 32BCode and reasoning | 32B | 1× $0.45/h | 1× $0.45/h | 1× $0.45/h |
| Mixtral 8x7BMixture of experts, high throughput | 46.7B | 2× $0.90/h | 1× $0.45/h | 1× $0.45/h |
| Llama 3.3 70BThe common production baseline | 70B | 4× $1.80/h | 2× $0.90/h | 1× $0.45/h |
| Qwen2.5 72BMultilingual, long context | 72B | 4× $1.80/h | 2× $0.90/h | 1× $0.45/h |
| Mixtral 8x22BMixture of experts, large capacity | 141B | 8× $3.60/h | 4× $1.80/h | 2× $0.90/h |
| Llama 3.1 405BLargest widely used open model | 405B | over 8× | 8× $3.60/h | 4× $1.80/h |
Real usage depends on context length, batch size and the serving engine: a long context can add tens of gigabytes of KV cache. Treat the table as the floor, not the ceiling — and if a job is close to the limit, take the next node size or quantise one step further.
What a dollar buys on an H100 PCIe.
Ranked 19 of 21 data-center models in this catalogue on cost per gigabyte of GPU memory.
GPU memory you get for $1.00 of spot time, per hour.
Bandwidth decides token throughput far more often than raw FLOPS.
One GPU on spot, billed per minute, storage excluded.
Against the rest of the market
20 providers · September 2026- Median on-demand elsewhere
- $2.68Our on-demand is 74% below it
- Cheapest on-demand seen
- $1.49Our on-demand $0.69 is 54% below it
- Cheapest spot seen
- $1.33Our spot $0.45 is 66% below it
- 24 hours on one GPU
- $10.80 on spot$16.56 on-demand · $35.76 at the cheapest rate found elsewhere
Index built from published prices for the same GPU across 137 providers (public price index (getdeploying.com)), September 2026. How the guarantee is enforced.
What people run on an H100 PCIe.
Training and full fine-tuning
Enough memory for optimiser states and activations on models a smaller card can only run in inference.
Serving large models
A 70B model at FP8 across two GPUs, with room for a long context window.
Partitioned serving (MIG)
The GPU can be split into isolated instances, each with its own memory and compute slice.
Anything that can checkpoint
On the spot tier this model is 35% below its own on-demand price, with a 2-minute notice before a reclaim.
NVIDIA H100 PCIe, as published by NVIDIA.
Taken from the vendor datasheet. Where a figure is not published, the row is absent rather than estimated.
- Vendor
- NVIDIA
- Architecture
- Hopper (2022)
- Segment
- Datacenter
- Memory
- 80 GB HBM2e
- Memory bandwidth
- 2,000 GB/s
- Board power
- 350 W
- Form factor
- PCIe dual-slot
- Interconnect
- PCIe Gen5 x16; NVLink bridge (3 bridges) 600 GB/s to one adjacent card
- Multi-instance (MIG)
- Supported by the GPU
- 8-bit float (FP8)
- Supported by the architecture
Source: www.nvidia.com · not confirmed on a vendor document
Software that runs on it
8 templatesPre-built environments, pulled on the node before you land on it. Or bring any OCI image from a public or private registry.
On the spot tier
<5% reclaimed · 30 d- 2 minutes of notice on the metadata endpoint, a webhook and the console
- The instance is stopped, not deleted — disk and IP stay attached
- Auto-relaunch on the next free H100 PCIe, or switch the same disk to on-demand
- You save $5.76 a day per GPU against on-demand
If the H100 PCIe is not the right fit.
Same memory or more, 58% cheaper on spot
H100 NVL 94 GB HBM3 · Hopper $0.49/h spotNext step up in memory: 94 GB instead of 80 GB
H200 SXM 141 GB HBM3e · Hopper $0.75/h spotSame Hopper generation
H200 NVL 141 GB HBM3e · Hopper $0.69/h spotSame Hopper generation
Or put two of them side by side: H100 PCIe vs RTX PRO 6000 · H100 PCIe vs H100 NVL · H100 PCIe vs H200
H100 PCIe — the questions that come up
How much does an H100 PCIe cost per hour?
Spot is $0.45 per GPU-hour and on-demand is $0.69, both billed per minute — every started minute costs the hourly price divided by 60, so $0.0075 on spot. Reserved capacity is $0.49 per GPU-hour on a monthly commitment. Storage and egress are not included in that rate: persistent disks are $0.08 per GB-month and there are no egress fees.
How many GPUs can I put in one instance?
Node sizes are 1×, 2×, 4×, 8× — up to 8 H100 PCIe GPUs in a single virtual machine, with CPU, RAM and local NVMe scaled with the GPU count. The price per GPU is identical at every node size.
Can it run a 70B model?
Yes. At FP8, a 70B model needs about 70 GB for the weights, so it fits on 2× H100 PCIe ($0.90 per hour on spot) with roughly 20% of the memory left for activations and the KV cache.
Which regions have it?
Available now in US East (Virginia), US West (Oregon), EU Central (Frankfurt). Prices are identical in every region; pick the one closest to your data.
How often is a spot H100 PCIe reclaimed?
Over the trailing 30 days, <5% of spot instances on this model were reclaimed. You get a 2-minute notice on the metadata endpoint, through a webhook and in the console; the instance is then stopped, never deleted, and the persistent disk stays attached.
Anything else about this model? Ask an engineer — the same people run the nodes.
An H100 PCIe for $0.45 an hour, running in under 60 seconds.
Pay as you go — no contracts, no minimum commitment. Add credit, launch, stop whenever you want.
Billed per minute from the moment the instance is reachable. Minimum credit $40, no subscription. 96 H100 PCIe GPUs available right now.