Kubernetes Kapsule
Orchestrate your containerized AI workloads with our Managed Kubernetes service.
GPU instances
Train, fine-tune, and serve your AI models on the market’s most powerful GPUs with guaranteed data residency within Europe.
Performance without compromise
Elite GPU performance
Access the latest GPU architectures and get the high-performance compute required for training LLMs and Generative AI.
True digital sovereignty
Your training datasets, models, and user data are hosted in Europe and completely protected from extraterritorial laws.
Predictable, transparent pricing
Eliminate the CAPEX barrier and the fear of hidden costs with simple, pay-as-you-go billing with no egress or lock-in.
Engineered for every AI workload
Choose the right GPU-as-a-Service for your specific AI needs, from model training to real-time inference.
B300-SXM
Based on NVIDIA’s latest Blackwell architecture, the B300-SXM GPU Instances push the boundaries of performance with increased NVIDIA NVLink bandwidth and the highest VRAM available to achieve complex reasoning.
H100-SXM
The NVIDIA H100-SXM Instances deliver superior performance and multi-GPU scalability thanks to higher power limits and advanced NVLInk (900 GB/s GPU-GPU communication).
H100 PCIe
The NVIDIA H100 PCIe Instances deliver versatile compute and rapid model fine-tuning speeds thanks to the Hopper architecture and built-in Transformer Engine (up to 30x faster model serving vs. previous generations).
L4
The L4 GPU Instance is a versatile, cost-effective, and energy-efficient solution for high throughput and low latency acceleration for video, AI, visual computing, graphics, and more.
L40S
Combining powerful AI compute with best-in-class graphics and media acceleration, the L40S GPU Instance is built to power workloads ranging from generative AI and LLM inference to 3D graphics, rendering, and video.
Optimize your AI infrastructure
Build and deploy on a platform designed for speed, transparency, and responsibility.
Scale instantly
Quickly spin up on-demand GPU Instances via our console, CLI, or Terraform.
Transfer data without penalty
We ensure predictable billing by eliminating hidden costs for moving data in or out of your infrastructure.
Run your models responsibly
Monitor the environmental impact of your AI models directly with our Environmental Footprint Calculator.
Estimate your GPU Instances cost
Choose your plan
Estimated cost
| Option and value | Price |
|---|---|
| ZoneParis 2 | |
| Instance1x | 0€ |
| Volume10GB | 0€ |
| Flexible IPv4No | 0€ |
A complete ecosystem for GPU for AI
Object Storage
Store massive training datasets securely and access them from your GPU Instances.
Managed Inference
Deploy AI models in a dedicated inference infrastructure.
Frequently asked questions
What is a GPU?
Graphical Processing Units (GPU) are specialized hardware originally designed for rendering graphics in video games and other 3D applications. However, their massively parallel architecture makes them an essential GPU for AI, powering complex tasks like deep learning, massive machine learning, data processing, scientific simulations, and more.
What is a cloud GPU Instance?
This refers to a virtual environment providing remote access to high-performance hardware over the internet. Utilizing a GPU in cloud computing allows you to leverage massive power without the burden of managing physical infrastructure.
How quickly can I get started?
How am I billed for GPU Instances consumption?
You can opt for a “pay as you go” model (GPUs are billed per minute), paying only for what you consume. This approach gives you the flexibility to provision resources and delete resources when needed.
Which operating system images are available?
GPU OS 13 is a specialized OS image based on Ubuntu Noble GPU OS 13 (Nvidia), optimized for GPU-accelerated workloads. It comes pre-installed with the NVIDIA driver, Docker, and NVIDIA's Container Toolkit, providing an environment for running containerized applications. This image is designed to work efficiently with NVIDIA NGC container images, enabling rapid deployment of GPU-accelerated applications, such as machine learning and data processing workloads.
Can I use GPU Instances with Kubernetes?
Yes, you can use GPU instances with our Kapsule, our managed K8S product. Kapsule will automatically manage the setup of the GPU thanks to the NVIDIA GPU Operator. Learn more about https://www.scaleway.com/en/docs/kubernetes/how-to/use-nvidia-gpu-operator/#how-to-get-the-gpu-operator-for-a-new-pool.
Is your GPU offer HDS (Hébergeur de Données Santé) compliant?
Yes, all of our GPU Instances are HDS compliant.
Do you offer Service Level Agreements (SLA)?
Yes, we offer a 99.5% SLA on all GPU Instances. Learn more here.
What type of support is available?
Scaleway offers 4 different support plans to match your needs: Basic, Advanced, Business, and Enterprise. You can find all the information, including pricing, here.
Which GPU is best suited for large-scale AI and high-performance computing?
For large-scale AI and high-performance computing (HPC) tasks, our AI Supercomputers deliver exceptional computational power and are optimized for intensive parallel processing. Check out our dedicated page for more information.
How to launch a GPU Instance using Docker?
To access the GPU with Docker on Scaleway GPU instances, use Docker with NVIDIA Container Toolkit installed to enable GPU support in containers. Make sure to deploy your container with the --gpus flag to access the GPU resources. Learn more here.
Can I use Savings Plans for GPUs?
No, GPUs aren't eligible to Savings Plans.
Can I switch easily to another GPU type ?
Yes, you can easily migrate your GPU. Learn more here.