The NVIDIA H100 PCIe Tensor Core GPU is joining Scaleway

Be the first to use the next-generation GPU

80GB of VRAM (HBM2e)

GPU memory sized to execute large AI models and datasets with 80GB of VRAM (HBM2e).

Tensor Core 4th generation

6x faster than A100 Tensor Core.

30x faster

Transformer Engine for 30x faster AI inference speedup on language models compared to the prior generation A100.

7x secure tenants

2nd generation Multi-Instance GPU for up to 7x secure tenants.

Sign up for the waiting list

The best engine for AI Startups

  • Speeding up training, iteration and go to production
    • Accelerate your inference workloads thanks to new data formats like Int4.
    • Fine-tune Transformers Models and train larger models faster thanks to the Tensor Cores 4th generation coupled with the new data format FP8.
    • Optimize your workloads by maximizing GPU utilization thanks to MIG (multi-instance GPU) partitioning.

Get started with our available GPU products

GPU 3070

€0.98 ex. VAT/hour

Powerful instance based on dedicated NVIDIA® GeForce® RTX 3070 offers excellent price/performance.

GPU Render

€1.221 ex. VAT/hour

Dedicated Tesla P100s for all your Machine Learning & Artificial Intelligence needs.

Outstanding features for your GPUs

High-end specifications

We paired one of the best processors – AMD EPYC™ 7003 Series with a world-class NVIDIA GeForce® RTX 3070 for your intensive workload.

API, CLI & SDK

We designed a suite of tools to support developers in the deployment, configuration, orchestration and automation of infrastructures, whatever their size and type.

Your own dedicated GPU

Scaleway GPU-3070 Instances come with your own dedicated GPU powered by NVIDIA CUDA®.

Scaleway ecosystem integration

Make your life easier by leveraging the Scaleway cloud services ecosystem with our Block Storage, Object Storage, Container Registry, and Kubernetes Kapsule.

Kubernetes-ready ecosystem

Deploy your Kubernetes nodes simply and easily with Kubernetes Kapsule, or accelerate your Docker containers with the NVIDIA® Container Toolkit to save valuable time.

Private Networks

Private Networks is compatible with GPU Instances at no extra cost. Privates Networks enables you to create your own layer-2 private networks between your Development Instances, General Purpose Instances and GPU Instances.