On Demand Cluster

Rent an On Demand Cluster — for a week, a month or several months — to speed up training requiring multiple nodes of aggregate GPU memory

Got a request? We’re all ears!

From 16 to 504 GPUs to support your development

Reserve the Cluster fitted for your need — from 16 to 504 GPUs — to secure your access to efficient NVIDIA H100 Tensor Core GPUs.

Fast Networking and GPU-GPU communication for distributed training

NVIDIA HGX H100 with NVlink & Spectrum-X Network accelerates the key communication bottleneck between GPUs and is one of the top solution on the market to run distributed training.

Private and secured environment

NVIDIA Spectrum-X —the latest Networking technology developed by NVIDIA— enables us to build multi-tenant clusters hosted in the same adiabatic Data Center.

Boost Innovation Sustainably: 100% Renewable Energy, 50% Less Power
DC5, in PAR2 region, is one of Europe's greenest data centers, powered entirely by renewable wind and hydro energy (GO-certified) and cooled with ultra-efficient free and adiabatic cooling. With a PUE of 1.16 (vs. the 1.55 industry average), it slashes energy use by 30-50% compared to traditional data centers.
Discover Scaleway's environmental commitments

Use cases

: The emerging class of trillion parameter AI models requires months to train, even on supercomputers. A GPU Cluster On-Demand enable user to compress this time and complete training within hours, thanks to high-speed, seamless communication between every GPU in a server cluster.

: Weights and parameters have become so large in recent LLMs that clusters of GPU (connected via high speed interconnect) are required to load the model into the memory for serving.
GPU Cluster On-Demand raison d’être is to annihilate this bottleneck and enable efficient inference of this gigantic models.