# Rent NVIDIA H200 and HGX H200 servers from $3,090 per month

> Rent a dedicated NVIDIA H200 (141 GB) for $3,090/month, about $4.23 per GPU-hour at 24/7, or an 8-GPU HGX H200 for $23,900. Flat price, paid in crypto.

Source: https://overgpu.com/gpus/nvidia-h200  
Updated: 2026-10-23  
Publisher: OverGPU (https://overgpu.com/)

**In short:** OverGPU rents the NVIDIA H200 by the month: a dedicated single H200 (141 GB HBM3e) for $3,090 per month, about $4.23 per GPU-hour at full use, or an 8-GPU NVIDIA HGX H200 server (8 × 141 GB HBM3e) for $23,900 per month, about $4.09 per GPU-hour. Both are in stock and delivered in under 10 minutes after payment. Payment is from a USD balance topped up in Bitcoin, Ethereum, Monero, USDT (TRC-20) or Solana, with no setup fee, and the plan can be cancelled before any monthly renewal.

## H200 configurations and monthly prices at OverGPU

| Configuration | Type | GPU memory | CPU | RAM | Storage | Network | Price per month (USD) | Per GPU-hour | Availability |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| [NVIDIA H200](https://overgpu.com/gpus/nvidia-h200) | Single GPU | 141 GB HBM3e | 24 vCPU | 240 GB | 2 TB NVMe | 25 Gbps | $3,090 | $4.23 | In stock, delivered in under 10 minutes after payment |
| [8× NVIDIA HGX H200](https://overgpu.com/gpus/nvidia-h200) | 8-GPU server | 8 × 141 GB HBM3e | 2× Xeon 8480+ · 112 cores | 2 TB DDR5 | 8× 3.84 TB NVMe | 8× 400G ConnectX-7 | $23,900 | $4.09 | In stock, delivered in under 10 minutes after payment |

Per GPU-hour = monthly price ÷ number of GPUs ÷ 730 hours, shown only to compare with hourly offers: every plan is billed as one flat monthly price.

## NVIDIA H200 specifications (vendor-published)

| Specification | Value | Note |
| --- | --- | --- |
| Architecture | Hopper | NVIDIA |
| GPU memory | 141 GB HBM3e |   |
| Memory bandwidth | 4.8 TB/s |   |
| FP4 | — | Not supported |
| FP8 | 1,979 TFLOPS | 3,958 TFLOPS with sparsity |
| BF16 / FP16 | 989 TFLOPS | 1,979 TFLOPS with sparsity |
| GPU interconnect | NVLink 4 · 900 GB/s | per GPU |
| Form factor | SXM module · HGX H200 |   |
| Max power | 700 W |   |

## What the H200 is best at

- **70B inference in FP8:** About 70 GB of weights leaves roughly 50 to 60 GB for KV cache on a single GPU.
- **Long-context serving:** A 128k-token context for a 70B model fits next to its FP8 weights on one H200.
- **Fine-tuning:** QLoRA of 70B models on one GPU, and full fine-tunes of 30B-class models on the 8-GPU server.
- **Large models on one node:** 1.1 TB of HBM3e across eight GPUs holds a 671B mixture-of-experts model in FP8.

## When another GPU fits better

- [Rent NVIDIA H100 SXM servers from $2,190 per month](https://overgpu.com/gpus/nvidia-h100): If your models fit in 80 GB with room for KV cache, the H100 has the same vendor-listed compute for $2,190 per month.
- [Rent NVIDIA B200 and HGX B200 servers from $4,590 per month](https://overgpu.com/gpus/nvidia-b200): When you need 180 GB per GPU, FP4, or the higher Blackwell compute figures NVIDIA lists for training.
- [Rent an 8-GPU AMD MI355X server for $19,900 per month](https://overgpu.com/gpus/amd-mi355x): If your stack runs on ROCm and you want 288 GB per GPU in an 8-GPU server for $19,900 per month.

## Overview

The main reason to rent NVIDIA H200 capacity is memory: 141 GB of HBM3e per GPU, enough to serve a 70B model in FP8 on a single card with room to spare. We offer the H200 in stock as a single GPU in a dedicated virtual machine at $3,090 per month, our most popular single-GPU plan, and as a bare-metal HGX H200 server with eight GPUs at $23,900 per month.

## What the H200 adds to the H100

The H200 uses the same Hopper GPU design as the H100. NVIDIA lists identical compute for both, 1,979 TFLOPS of dense FP8 and 989 TFLOPS of dense BF16, the same NVLink 4 at 900 GB/s per GPU and the same 700 W SXM form factor. The difference is memory: 141 GB of HBM3e instead of 80 GB of HBM3 (76% more), at 4.8 TB/s instead of 3.35 TB/s (43% more).

That matters most for inference. To generate each token, the GPU reads the model weights from memory, so bandwidth sets the ceiling on decode speed, and every gigabyte not taken by weights can hold KV cache for more concurrent requests or longer contexts. For compute-bound work, such as training a model that already fits comfortably in 80 GB, the gap narrows and the H100 may be the better value. Our [H100 vs H200 comparison](https://overgpu.com/compare/h100-vs-h200) lays out both cases.

## What fits on one H200, and on eight

Weights take roughly 2 bytes per parameter in BF16 and 1 in FP8; add KV cache and 10–30% headroom.

- **70B in FP8 on one H200:** about 70 GB of weights, leaving roughly 50–60 GB for KV cache after headroom. That headroom is what makes a single H200 practical for a production 70B endpoint.
- **Long contexts:** a Llama-style 70B model with grouped-query attention (80 layers, 8 KV heads of dimension 128) needs about 0.33 MB of KV cache per token in BF16. One 128k-token context is therefore about 43 GB, which fits next to the FP8 weights on an H200 but not on an 80 GB GPU.
- **70B in BF16:** about 140 GB, too tight for one H200. Use two GPUs of the HGX H200 with tensor parallelism, or a 180 GB B200.
- **HGX H200, 1.1 TB in total:** a 405B-class model in FP8 (about 405 GB) leaves roughly 700 GB for KV cache, and a 671B mixture-of-experts model in FP8 (about 671 GB) still fits with around 450 GB to spare.

For fine-tuning, QLoRA on a 70B model (about 35 GB of 4-bit weights) runs comfortably on one H200, while LoRA on 70B in BF16 does not fit one GPU because the frozen weights alone take about 140 GB. Full fine-tuning with Adam needs roughly 16–20 bytes per parameter before activations: about 512–640 GB for a 32B model, which the HGX H200 handles, but 1.1–1.4 TB for a 70B model, which exceeds its HBM3e unless you offload optimizer state to the 2 TB of system RAM. Try other sizes in our [VRAM calculator](https://overgpu.com/tools/vram-calculator).

## Single H200 or an HGX H200 server?

The single H200 comes with 24 vCPUs, 240 GB of RAM, 2 TB of NVMe and a 25 Gbps port. It fits teams that serve one model per GPU and scale by adding replicas, or that fine-tune with LoRA and QLoRA.

The HGX H200 links eight GPUs in one NVLink domain at 900 GB/s per GPU, so you can split a model with tensor parallelism or shard training state across all eight without touching the network. The server has two Xeon 8480+ processors (112 cores in total), 2 TB of DDR5, eight 3.84 TB NVMe drives and eight 400G ConnectX-7 adapters. At eight GPUs it also costs less than single plans: $23,900 per month, against $24,720 for eight single H200s.

## Software: a mature Hopper stack

Hopper GPUs have been widely deployed since 2023, so FP8 support in vLLM, SGLang and TensorRT-LLM is well established. Your server arrives with root access over SSH, Ubuntu 22.04 or 24.04, and NVIDIA drivers, CUDA, Docker and the NVIDIA Container Toolkit preinstalled; Debian or Rocky Linux are available on request. Serving a 70B model from an FP8 checkpoint, with an FP8 KV cache, on one H200 looks like this:

```
vllm serve <your-70b-fp8-checkpoint> --kv-cache-dtype fp8 \
  --max-model-len 131072 --host 127.0.0.1 --port 8000
```

Storing the KV cache in FP8 halves its size, which roughly doubles the context or concurrency that fits in the same memory. Our guide to [self-hosting LLM inference with vLLM](https://overgpu.com/blog/self-host-llm-inference-vllm) covers production settings, and [choosing a GPU for LLM inference](https://overgpu.com/blog/best-gpu-for-llm-inference) compares the options.

## H200 price per month and per GPU-hour

- **Single H200:** $3,090 per month, about $4.23 per GPU-hour
- **HGX H200, 8 GPUs:** $23,900 per month, about $4.09 per GPU-hour

The hourly figure is the monthly price divided by the number of GPUs and by 730 hours, so it holds only if the GPU runs all month. A 24/7 inference endpoint does, and there a flat $3,090 replaces a meter. An endpoint that only runs during business hours, about 220 hours a month, would cost about $14.05 per used hour on a monthly plan, and hourly billing is then likely the better choice. The general rule: divide our monthly price by the hourly rate you would pay elsewhere, and monthly wins above that number of hours. Our [monthly vs hourly comparison](https://overgpu.com/compare/monthly-vs-hourly-gpu-pricing) shows the math in detail.

Prices have also been moving. OverGPU's October 2026 survey of public price lists found the median on-demand H200 price per GPU-hour up roughly 27% over the past year. A monthly plan fixes your price for each paid month. It includes outbound traffic with no egress fees (20 TB per month for the single GPU, 50 TB for the 8-GPU server, under [fair use](https://overgpu.com/help/bandwidth-and-egress)), a public IPv4 address and an IPv6 /64, DDoS mitigation, 24/7 support and the 99.9% availability SLA. There is no setup fee.

## How to rent NVIDIA H200 servers

Choose the plan you want on the [deployment page](https://overgpu.com/deploy), set up the OS image, optional preinstalled software and your login (root password or SSH key), and create a free account or log in in the same form. You then pay the first month from your prepaid balance, which you top up in BTC, ETH, XMR, USDT on TRON or SOL at an exchange rate locked for a limited time, with up to 50% bonus credit; card payments are coming soon. Root SSH access is delivered in under 10 minutes after payment, with the IP and SSH details in your dashboard. Plans renew monthly from your balance with an email reminder, and you can cancel any month without a fee. For volume pricing on several servers or a 6–12 month term, customers with an active server can ask by ticket from their dashboard.

## Frequently asked questions

### What is the H200 price per month?

A single H200 is $3,090 per month and an 8-GPU HGX H200 server is $23,900 per month, about $4.23 and $4.09 per GPU-hour at full use. The price is flat, with no setup fee and bandwidth included.

### How is the H200 different from the H100?

NVIDIA lists the same FP8 and BF16 compute for both. The H200 has 141 GB of memory instead of 80 GB and 4.8 TB/s of bandwidth instead of 3.35 TB/s, which helps memory-bound work such as LLM inference with large batches or long contexts.

### Can I run a 70B model on a single H200?

In FP8, yes: the weights take about 70 GB and leave roughly 50 to 60 GB for KV cache after headroom. In BF16 the weights alone take about 140 GB, which is too tight for one H200.

### Is the H200 in stock?

Yes, the single H200 and the HGX H200 are both in stock at the time of writing. Both are delivered in under 10 minutes after payment.

### Can I cancel my H200 rental?

Yes, any month, with no cancellation fee. Your server runs until the end of the paid month, and its disks are securely erased afterwards.

### Can several single H200 plans talk to each other?

Yes, through private networking between your servers, available on request. For jobs that need NVLink between GPUs, take the HGX H200 instead.

## Related pages

- [H200 vs B200: Hopper or Blackwell for your next GPU server?](https://overgpu.com/compare/h200-vs-b200): 141 vs 180 GB, 4.8 vs 8 TB/s, FP4 and NVLink 5. See what fits, what $1,500 more a month buys and when the H200 is the better buy.
- [MI355X vs H200: double the memory or the proven stack?](https://overgpu.com/compare/mi355x-vs-h200): Twice the memory per GPU and native FP4 against the mature Hopper and CUDA stack. Compare what fits, interconnect and monthly cost.
- [GPU servers for LLM inference, running 24/7 at a flat price](https://overgpu.com/solutions/inference): KV cache sizing, FP8 and FP4, and 20–50 TB of bandwidth included. See the sizing math.
- [The right GPU for LLM fine-tuning, from QLoRA to full fine-tunes](https://overgpu.com/solutions/fine-tuning): Memory math for QLoRA, LoRA and full fine-tunes of 8B to 70B models on H100, H200 and B200. Check what fits first.
- [Single GPU server rental: a whole GPU in a dedicated VM](https://overgpu.com/products/single-gpu-servers): One whole GPU in a dedicated cloud VM, from RTX 5090 to B300. Root access, 20 TB bandwidth, monthly billing, delivered in under 10 minutes.
- [Serve an LLM with vLLM](https://overgpu.com/docs/pytorch-vllm): Set up Python with uv, install PyTorch and vLLM, test the OpenAI-compatible API and run it as a service.