Sesterce
The European AI Factory company - sovereign by design, renewable by default
About Sesterce
Sesterce is a European cloud computing company providing sovereign, energy-efficient GPU infrastructure for AI training, inference, and HPC workloads. It operates purpose-built AI factories across France with vertically integrated systems spanning energy, hardware, networking, and orchestration, giving organizations on-demand access to NVIDIA GPUs while maintaining data residency and supply-chain control within Europe.
Pros & Cons
Pros
- European data sovereignty with full control over data residency and energy origin
- Broad NVIDIA GPU selection from budget RTX cards to flagship B300
- Competitive on-demand pricing with no commitment required
- Renewable-powered, liquid-cooled infrastructure with strong PUE efficiency
- Fast cluster provisioning (under two hours) without ticket queues
- Self-service CLI, API, and Terraform support for automation
Cons
- Infrastructure is Europe-based, which may add latency for users in other regions
- Geared toward AI/ML and HPC workloads rather than general-purpose cloud needs
- Enterprise sovereign offerings require custom contracts
- Spot and reserved pricing benefits depend on capacity availability
Key Features
On-Demand GPU Cloud
Launch on-demand VMs and bare-metal servers in seconds with no commitment, across a wide range of NVIDIA GPUs from RTX 4090 to B300.
Latest NVIDIA Hardware
Access H100, H200, A100, L40S, B200, B300, and RTX Pro 6000 GPUs with 800 Gb/s InfiniBand connectivity for large-scale training.
Sovereign Compute
Maintain data residency, energy origin, and supply-chain control with infrastructure hosted entirely within Europe.
Sesterce OS Control Plane
Unified control plane supporting Slurm and Kubernetes that provisions clusters, orchestrates training, and serves inference from a single pane of glass.
Sub-2-Hour Provisioning
Go from signed contract to live cluster in under two hours with no ticket queues.
Liquid-Cooled Infrastructure
Energy-efficient liquid cooling at up to 150 kW per rack with a PUE of 1.12, powered by renewable energy by default.
Dedicated Resources
Dedicated hardware, network, and storage with no noisy neighbors for predictable performance.
Self-Service Tooling
CLI, API, and Terraform support for multi-cloud deployments and infrastructure-as-code workflows.
Real-Time Telemetry
Rack, node, GPU, and job-level monitoring via DCGMI exposed through APIs.
Flexible Billing Modes
Choose between on-demand, reserved commitment, and spot pricing to optimize cost for each workload.
Pricing
- RTX A4000 from $0.17/hr
- RTX 4090 from $0.66/hr
- RTX 5090 from $0.71/hr
- On-demand, no commitment
- A100 40GB from $1.50/hr
- H100 from $2.09/hr
- H200 from $3.78/hr
- Persistent volumes and bare metal
- B200 from $7.17/hr
- B300 from $8.70/hr
- 800 Gb/s InfiniBand
- Reserved commitment and spot options
- Dedicated AI factory capacity
- Sub-2-hour cluster provisioning
- Sovereign-hosted data stores and HSM key management
- Slurm and Kubernetes orchestration
Best For
Large-Scale AI Training
Train large models on clusters of H100, H200, or B200 GPUs connected over high-speed InfiniBand.
Production Inference
Serve AI models in production on the same fabric used for training via a unified API.
Sovereign AI for Regulated Industries
Run AI workloads with European data residency and supply-chain control for compliance-sensitive sectors.
HPC and Research Workloads
Run high-performance computing jobs with Slurm orchestration and dedicated resources.
Cost-Optimized Experimentation
Spin up affordable entry-level GPUs on-demand for prototyping and small-scale experiments.
Similar Tools
Vultr
High-performance cloud compute, GPU, and bare metal across 32 global data centers
RunPod
The end-to-end GPU cloud for AI workloads
Lambda
The superintelligence cloud for GPU compute and AI infrastructure
GauGAN2
NVIDIA's AI that turns text and sketches into photorealistic landscape images
Milvus
High-performance, cloud-native vector database built for scalable AI applications
MinIO
High-performance, S3-compatible object storage for AI and enterprise workloads