# Rack-scale GPU systems: GB300 NVL72 and Vera Rubin NVL72

> Rack-scale GPU systems: rent GB300 NVL72 from $7,990 per GPU-month on 12–36 month terms, or reserve Vera Rubin NVL72. See who needs a 72-GPU NVLink domain.

Source: https://overgpu.com/products/rack-scale-systems  
Updated: 2026-10-23  
Publisher: OverGPU (https://overgpu.com/)

## Key points

- **72 GPUs in one NVLink domain:** All 72 GPUs of a GB300 NVL72 share one NVLink 5 domain, with 130 TB/s of total NVLink bandwidth according to NVIDIA.
- **About 20 TB of HBM3e per rack:** 72 Blackwell Ultra GPUs with 288 GB of HBM3e each, plus 17 TB of LPDDR5X attached to 36 Grace CPUs.
- **Liquid-cooled:** NVL72 racks are liquid-cooled systems. You rent the capacity without building that cooling yourself.
- **12 to 36 month terms:** GB300 NVL72 starts at $7,990 per GPU per month, with your exact price confirmed before you pay.
- **Vera Rubin NVL72 reservations:** 72 Rubin GPUs with 288 GB of HBM4 each, 36 Vera CPUs and NVLink 6, with limited availability in late 2026.
- **Scoped with engineers:** Configuration, price and location are confirmed in writing before you pay.

## Recommended configurations

- [NVIDIA GB300 NVL72](https://overgpu.com/gpus/nvidia-gb300-nvl72): from $7,990 per GPU-month — One 72-GPU NVLink domain for frontier-scale training
- [NVIDIA Vera Rubin NVL72](https://overgpu.com/gpus/nvidia-vera-rubin-nvl72): price on request — Reserve Rubin capacity with 288 GB of HBM4 per GPU
- [8× NVIDIA HGX B300](https://overgpu.com/gpus/nvidia-b300): $39,900 per month — The same B300 GPU in 8-GPU servers, on monthly terms

## Overview

Rack-scale GPU systems turn a whole rack into one NVLink domain. In an NVIDIA GB300 NVL72, 72 Blackwell Ultra GPUs and 36 Grace CPUs are connected so that tensor, expert and pipeline parallel traffic can stay on NVLink across all 72 GPUs instead of crossing a network. We rent GB300 NVL72 on 12 to 36 month terms, and we are taking reservations for Vera Rubin NVL72.

## How a 72-GPU NVLink domain works

In an 8-GPU HGX server, NVLink connects eight GPUs. Anything beyond those eight crosses the cluster network, which offers far less bandwidth per GPU than NVLink. Large training and serving jobs are usually laid out around that boundary: tensor parallelism inside a server, data or pipeline parallelism between servers.

An NVL72 moves the boundary to 72 GPUs. NVIDIA publishes 130 TB/s of total NVLink bandwidth for the rack. Each GPU keeps 288 GB of HBM3e at about 8 TB/s, for about 20 TB of HBM3e per rack, and the 36 Grace CPUs add 17 TB of LPDDR5X. NVIDIA's vendor-published figures also put each GPU in the NVL72 at 5 PFLOPS of dense FP8, against 4.5 PFLOPS for a B300 in an HGX server.

The rack is liquid-cooled. Cooling at this density needs facility-level support, which is part of what you rent rather than build. The full vendor figures are on our [GB300 NVL72 page](https://overgpu.com/gpus/nvidia-gb300-nvl72).

## GB300 NVL72 rental: terms and pricing

GB300 NVL72 starts at $7,990 per GPU per month. For a full rack of 72 GPUs, that starting rate works out to $575,280 a month. Terms run from 12 to 36 months, unlike our servers, which you can rent month to month. Our team confirms each system in your dashboard with its configuration, price and location before you pay.

A 12-month minimum makes sense for a sustained program of work, such as a pre-training run followed by post-training and evaluation. If you are not sure your workload needs a 72-GPU domain, read the comparison below first, or start with an [HGX B300 server](https://overgpu.com/products/8-gpu-servers) on monthly terms.

## Vera Rubin NVL72 reservations

Vera Rubin NVL72 is the next NVIDIA rack-scale generation: 72 Rubin GPUs with 288 GB of HBM4 each, 36 Vera CPUs and NVLink 6, in a liquid-cooled rack. Availability is limited in late 2026, and reservations are open. There is no list price yet: the price comes with the confirmation of your reservation. In the notes of a [reservation request](https://overgpu.com/deploy), give the scale you need, the date you need it and the term you have in mind, and we come back with what we can deliver. Nothing is charged when you send it.

## Rack-scale GPU systems or several HGX servers?

Seventy-two B300 GPUs can also be nine HGX B300 servers. The right choice depends on how far your model has to be split.

| Option | GB300 NVL72 | 9× HGX B300 |
| --- | --- | --- |
| GPU memory | 72 × 288 GB HBM3e | 72 × 288 GB HBM3e |
| NVLink domain | 72 GPUs | 8 GPUs per server |
| Between NVLink domains | Not needed inside the rack | Private networking between servers |
| Dense FP8 per GPU (vendor) | 5 PFLOPS | 4.5 PFLOPS |
| Term | 12 to 36 months | Monthly; term pricing by ticket for active customers |
| Price | From $7,990 per GPU-month | $39,900 per server, about $4,990 per GPU-month |

Choose an NVL72 when tensor or expert parallelism has to span more than eight GPUs: very large mixture-of-experts models whose all-to-all traffic would otherwise cross the network, or serving setups whose weights and KV cache outgrow an 8-GPU server. Choose HGX servers when your parallelism fits inside eight GPUs and data parallelism handles the rest. You keep monthly flexibility and pay less per GPU, and we can link the servers with [private networking](https://overgpu.com/products/private-networking).

## How a rack-scale order works

1. [Send a reservation request](https://overgpu.com/deploy) from the deployment page, creating a free account or logging in on the same form, and describe the workload, scale, start date and term in the notes. Nothing is charged when you send it.
2. Our engineers scope the system from your request: configuration and location.
3. Rack-scale orders may require verification under our [verification policy](https://overgpu.com/legal/verification). We tell you exactly what is needed and why.
4. We confirm the configuration, price and location in writing, in your dashboard and by email, before you pay.

Scoping goes faster when you bring:

- the model size, the precision and the parallelism layout you plan, such as tensor, expert and pipeline sizes;
- how much data you need to move in, and where checkpoints should be copied;
- your software stack. Grace is an Arm CPU, so containers and Python dependencies need arm64 builds. Check custom packages before you commit to a term.

Support from engineers is available 24/7 for the whole term.

## Frequently asked questions

### What is an NVL72?

A rack-scale NVIDIA system with 72 GPUs and 36 CPUs, in which all 72 GPUs share one NVLink domain and exchange data over NVLink rather than a network. GB300 NVL72 uses Blackwell Ultra GPUs with Grace CPUs, and Vera Rubin NVL72 uses Rubin GPUs with Vera CPUs.

### How much does GB300 NVL72 rental cost?

From $7,990 per GPU per month, on terms of 12 to 36 months. The exact price for your configuration and term is confirmed in writing before you pay.

### Can I rent a GB300 NVL72 month to month?

No. GB300 NVL72 is rented on terms of 12 to 36 months. For monthly terms on the same B300 GPU, choose HGX B300 servers, alone or linked by private networking.

### Is Vera Rubin NVL72 available now?

Availability is limited in late 2026 and reservations are open. There is no list price yet: the price is confirmed in writing before you pay.

### Do I need a rack-scale system or several HGX servers?

If tensor or expert parallelism must span more than eight GPUs, a 72-GPU NVLink domain helps. If your parallelism fits within eight GPUs, several HGX servers linked by private networking keep monthly flexibility and a lower price per GPU.

## Related pages

- [GB300 NVL72 rental from $7,990 per GPU-month](https://overgpu.com/gpus/nvidia-gb300-nvl72): GB300 NVL72 rental from $7,990 per GPU-month (about $10.95 per GPU-hour) on 12–36 month terms: 72 Blackwell Ultra GPUs in one NVLink domain. Reserve now.
- [Rent NVIDIA Vera Rubin NVL72: reserve rack-scale capacity](https://overgpu.com/gpus/nvidia-vera-rubin-nvl72): 72 Rubin GPUs with 288 GB of HBM4 each and 36 Vera CPUs. Price on request; terms confirmed before any payment.
- [8-GPU server rental: bare-metal NVIDIA HGX and AMD Instinct](https://overgpu.com/products/8-gpu-servers): NVIDIA HGX B300, B200, H200, H100, AMD MI355X or 8x RTX PRO 6000. 50 TB bandwidth, monthly terms, delivered in under 10 minutes.
- [Private network between GPU servers](https://overgpu.com/products/private-networking): Link your OverGPU servers on an isolated network for data pipelines and multi-node jobs. See how it works.
- [GPU servers for LLM training on dedicated 8-GPU nodes](https://overgpu.com/solutions/llm-training): Bare-metal HGX B300, B200 and H200 nodes with NVLink and 400–800G networking, rented monthly. See the memory math first.
- [Dedicated GPU servers for enterprise, with a written SLA](https://overgpu.com/solutions/enterprise): Whole GPUs, bare-metal 8-GPU nodes, a 99.9% SLA with credits, VAT invoices and term pricing. Plan a pilot.