H100 SXM5 on Scaleway
Visit ScalewayScaleway's NVIDIA H100 SXM5 offering delivers enterprise-grade Hopper architecture GPUs with 80GB HBM3 VRAM, optimized for demanding AI, ML, and HPC workloads. As a leading European cloud provider, Scaleway emphasizes data sovereignty through its Paris and Amsterdam data centers, ensuring GDPR compliance and low-latency access for EU-based teams. This combination stands out via integration with the Nabu AI Supercomputer—a massive cluster of thousands of H100s enabling seamless scaling from single instances to exascale training. Key value propositions include per-hour billing for cost efficiency, strong environmental credentials with 100% renewable energy, and unified services like object storage, Kubernetes, and managed databases. Ideal for ML engineers prioritizing compliance, sustainability, and high-performance multi-GPU setups without vendor lock-in, it supports frameworks like PyTorch and TensorFlow out-of-the-box, accelerating inference and fine-tuning at scale.
Why NVIDIA H100 SXM5 on Scaleway?
Choose Scaleway for NVIDIA H100 SXM5 if EU data sovereignty is paramount, as instances run in French/Dutch regions with strict compliance. The Nabu Supercluster provides unique multi-node scaling up to thousands of GPUs via high-speed InfiniBand, complementing the H100's NVLink 900GB/s bandwidth for distributed training. Per-hour billing minimizes costs for bursty workloads, while integrated services (e.g., RKE2 Kubernetes, block/object storage) streamline deployments. Scaleway's renewable energy focus appeals to sustainability-conscious teams, and competitive pricing undercuts hyperscalers for long runs. This setup excels for regulated industries needing Hopper's Transformer Engine without transatlantic latency or data export risks.
Live Pricing
Real-time NVIDIA H100 SXM5 offers from Scaleway
| Provider | GPU Model | VRAM | Host Specs | Region | Price | Status | Action | |
|---|---|---|---|---|---|---|---|---|
 Scaleway | 4×NVIDIA H100 SXM5 80GB VRAM | 80GB | 64 vCPU 480GB RAM 5960GB Storage | Paris | $3.59/GPU/hr $14.37/hr total (4×) | Available | ||
Scaleway | 8×NVIDIA H100 SXM5 80GB VRAM | 80GB | 128 vCPU 960GB RAM 11921GB Storage | Paris | $3.61/GPU/hr $28.90/hr total (8×) | Sold Out | ||
Scaleway | 2×NVIDIA H100 SXM5 80GB VRAM | 80GB | 32 vCPU 240GB RAM 2980GB Storage | Paris | $3.72/GPU/hr $7.45/hr total (2×) | Available |
Performance Notes
On Scaleway, H100 SXM5 delivers NVIDIA-spec peaks: up to 4 petaFLOPS FP8 AI perf, 2 petaFLOPS FP16, with 80GB HBM3 at 3.35TB/s bandwidth. Nabu cluster offers 400Gbps HDR InfiniBand for multi-GPU/node scaling, supporting efficient NCCL all-reduce. Single instances suit prototyping; multi-node via Slurm excels for large models. NVMe storage options provide 10-30GB/s throughput. Real-world benchmarks show 80-95% scaling efficiency in PyTorch DDP. Network egress is solid at 100Gbps+, but exact instance interconnects vary—confirm via docs. No public HGX specifics yet; expect strong but test for your workload.
A major European cloud provider emphasizing data sovereignty and integrated services.
Best For
Unique Features
- Nabu AI Supercomputer
- Strong environmental credentials
VRAM
80GB
Architecture
Hopper
Tier
enterprise
Platform Features
Getting Started
Getting started with Scaleway's H100 SXM5 is straightforward via their intuitive console. New users sign up for a free account, fund via credit card/IBAN, and launch GPU instances in minutes. Pre-configured images with CUDA 12.x, NVIDIA drivers, and Docker support rapid ML workflows. Access Nabu queue for larger jobs post-setup.
Steps
- 1Create a Scaleway account at console.scaleway.com and add payment method.
- 2Navigate to Instances > GPU, select H100 SXM5 configuration (e.g., nabu-large).
- 3Choose region (Paris-dc1 preferred for Nabu), image (Ubuntu/CUDA), and storage.
- 4Configure SSH keys/network, review per-hour pricing, then launch instance.
- 5SSH in, pull NGC containers (e.g., nvcr.io/nvidia/pytorch), and start training.
Pro Tips
- Leverage Nabu House slurm queue for multi-node jobs—submit via sacctmgr after approval for cost-effective scaling.
- Use Scaleway's Elements Object Storage for datasets; mount via s3fs for seamless H100 data pipelines.
- Monitor GPU utilization with DCGM and Prometheus; enable auto-scaling for variable ML inference loads.
Frequently Asked Questions
What is Scaleway's billing model for NVIDIA H100 SXM5?▾
Scaleway bills per-hour for GPU instances including NVIDIA H100 SXM5. Hourly billing means you pay for full hours even if your job completes mid-hour. Plan your workloads accordingly to maximize cost efficiency.
Does Scaleway offer spot instances for NVIDIA H100 SXM5?▾
No, Scaleway does not currently offer spot instances for NVIDIA H100 SXM5. All instances are billed at on-demand rates. However, they do offer reserved instances for committed usage, which can provide significant discounts for long-term workloads.
Does Scaleway offer reserved instances for NVIDIA H100 SXM5?▾
Yes, Scaleway offers reserved instance pricing for NVIDIA H100 SXM5, which can provide significant discounts (typically 20-40% off on-demand rates) for committed usage periods. Reserved instances are ideal for predictable, long-running workloads like production inference services, ongoing training pipelines, or development environments that run continuously. Contact Scaleway for current reserved pricing and commitment terms.
Related Pages
Rent NVIDIA H100 SXM5
Atlantic.net vs Scaleway: GPU Cloud Comparison
AWS vs Scaleway: GPU Cloud Comparison
Cirrascale vs Scaleway: GPU Cloud Comparison
NVIDIA B300 SXM6 on Scaleway - Pricing & Availability
NVIDIA H100 PCIe on Scaleway - Pricing & Availability
NVIDIA L4 on Scaleway - Pricing & Availability
NVIDIA GeForce RTX 3070 on Scaleway - Pricing & Availability