Sesterce
Sovereign European AI factories and on-demand GPU cloud
Last reviewed Mar 31, 2026
Sesterce is a French AI infrastructure company running vertically integrated "AI factories" — owning the stack from energy supply and liquid-cooled facilities through NVIDIA silicon, InfiniBand fabric and its own orchestration layer — alongside a self-serve GPU cloud with on-demand VMs, bare-metal servers and dedicated clusters.
Available GPUs
Hourly on-demand pricing. Click column headers to sort.
Prices last updated: September 28, 2026
Sesterce pricing by GPU
Configurations, price rank, and alternatives for one GPU at a time.
Pros & Cons
Advantages
- Broad on-demand catalog running from Blackwell B300 and H200 down to previous-generation NVIDIA cards, plus Intel Gaudi 2
- On-demand, spot, and reserved billing modes on the same clusters
- VM and bare-metal deployment with single- and multi-GPU (up to 8x) configurations
- European data residency across French sites, with hydro, nuclear, and solar energy under long-term PPAs
- Slurm- and Kubernetes-native orchestration with rack-, node-, and GPU-level telemetry
- API- and CLI-first, with a published REST API reference
Limitations
- Infrastructure concentrated in France, with no North American or Asia-Pacific regions
- Rack-scale GB300 NVL72 and B200 capacity is contracted as reserved AI factory capacity rather than sold from the self-serve catalog
- The newest and largest accelerators (B300, H200, Gaudi 2) are only sold as 8-GPU nodes, with no smaller self-serve size
- Public documentation lags the console: it omits several GPU types the compute page offers and its own pricing table has not been updated in about a year
- Credit-based billing requires pre-funded balance
Key Features
Vertically Integrated AI Factories
Each site is operated as a single system, owning the hand-off between energy, facility, silicon, network fabric, and orchestration
Sovereign European Infrastructure
French sites offering data residency, controlled energy origin, and supply-chain control, drawing on 2 GW of hydro, nuclear, and solar capacity secured under long-term PPAs
SoftBank Joint Venture at Bosquel
A joint venture with SoftBank Group to develop a 1 GW AI data center campus at Bosquel in Hauts-de-France
Sesterce OS Control Plane
Single console and API to provision clusters, orchestrate training, serve inference, and meter usage across every Sesterce site, Slurm and Kubernetes native
High-Density Liquid Cooling
Purpose-built Tier III+ facilities with direct-to-chip liquid cooling at up to 150 kW per rack, rated at PUE 1.12 under full AI load
InfiniBand Fabric
Non-blocking NVIDIA Quantum InfiniBand at 800 Gb/s with rail-optimized topology and native RDMA
Zero-Touch Provisioning
Dedicated clusters go from signed contract to live hardware in under 2 hours, with single-tenant hardware, network, and storage
API and CLI First
Launch, scale, and destroy clusters from the CLI or REST API; every console action has an API equivalent
Multiple Deployment Options
Choose on-demand VMs, bare-metal servers, or dedicated clusters based on your workload requirements
Compute Services
Compute
On-demand VMs and bare-metal servers launched with no commitment
Clusters
Dedicated multi-node GPU clusters ranging from around 100 to 15,000 GPUs
AI Inference
Managed inference instances for serving models alongside GPU compute
Storage
Persistent volumes attachable to compute instances and reusable across runs
Pricing Options
| Option | Details |
|---|---|
| On-Demand | Pay-per-second usage with no commitment on VMs and bare-metal servers |
| Spot | Discounted interruptible capacity, filterable as a spot offer in the compute catalog |
| Reserved Capacity | Committed multi-megawatt cluster capacity contracted for frontier-scale workloads |
| Credit-based Billing | Pre-funded account balance system for streamlined resource management |
Availability & Support
Regions
French footprint: AI factory sites at Bosquel (Hauts-de-France) and Valence, plus a Paris cloud region hosted at Equinix; several on-demand GPU types are offered in more than one region
Support
Documentation and API reference, self-service console with help and requests, REST API and CLI, contact sales for reserved capacity
Getting Started
- 1
Create Account
Sign up for Sesterce Cloud platform access
- 2
Add Credits
Fund your account with credits for pay-as-you-go billing
- 3
Launch Instance
Choose your GPU type and configuration, launch in just a few clicks or via the CLI and API
Frequently Asked Questions
What GPU types does Sesterce offer?
Sesterce offers various GPU types including RTX A5000, RTX A5000, RTX A5000, RTX A6000, RTX A6000, RTX A6000, A100 SXM, A100 SXM, A100 SXM, A100 SXM, H100 SXM, H100 SXM, H100 SXM, H200, L40, L40, L40, L40, A10, RTX 4000 Ada, RTX 4090, A40, RTX 6000 Ada, RTX 6000 Ada, RTX 6000 Ada, RTX A4000, RTX A4000, L40S, L40S, L40S, L40S, L40S, RTX 5090, RTX 5090, RTX 5090, Tesla V100, Tesla V100, Tesla V100, A16, L4, L4, L4, RTX PRO 6000, RTX PRO 6000. Check the pricing table above for current availability and pricing.
How do I get started with Sesterce?
Create Account, Add Credits, Launch Instance
What are Sesterce's main advantages?
Sesterce's main advantages include: Broad on-demand catalog running from Blackwell B300 and H200 down to previous-generation NVIDIA cards, plus Intel Gaudi 2, On-demand, spot, and reserved billing modes on the same clusters, VM and bare-metal deployment with single- and multi-GPU (up to 8x) configurations, European data residency across French sites, with hydro, nuclear, and solar energy under long-term PPAs, Slurm- and Kubernetes-native orchestration with rack-, node-, and GPU-level telemetry, API- and CLI-first, with a published REST API reference.
What are Sesterce's limitations?
Sesterce's main limitations include: Infrastructure concentrated in France, with no North American or Asia-Pacific regions, Rack-scale GB300 NVL72 and B200 capacity is contracted as reserved AI factory capacity rather than sold from the self-serve catalog, The newest and largest accelerators (B300, H200, Gaudi 2) are only sold as 8-GPU nodes, with no smaller self-serve size, Public documentation lags the console: it omits several GPU types the compute page offers and its own pricing table has not been updated in about a year, Credit-based billing requires pre-funded balance.
Compare Providers
Find the best prices for the same GPUs and models from other providers
