1:23
Damien Patriarche
President, Architect
Patriarche
“
Sesterce builds infrastructure ready for long-term AI evolution.
01
Compute. Made in Europe.
Sovereign AI
infrastructure,
end to end.
Featured site
Sesterce SoftBank Bosquel
Trusted by
NVIDIASoftBankSociete GeneraleCredit AgricoleF5UMA RoboticsBNP ParibasPatriarche
Our AI Factories
Three sites. One sovereign grid.
Purpose-built French AI factories and sovereign cloud sites. Liquid-cooled, renewable-powered.
Sovereign AI compute,
on the continent where you train.
Data residency, energy origin, and supply-chain control matter.
The factory
From watt to token,
end to end.
Every Sesterce factory is a single vertically-integrated system. We own the hand-off between every stage.
Layer 01
Energy
2 GW of clean energy under control in Europe - hydro, nuclear and solar, secured on long terms PPAs.
Green First
Layer 02
Facility
Purpose-built, direct-to-chip liquid-cooled, Tier III+ certified. PUE 1.12 under full AI load.
150 kW / rack
Layer 03
Silicon
NVIDIA GB300, H200 and B200 - latest generation, Blackwell-ready on day one.
NVIDIA GB300 NVL72
Layer 04
Fabric
Non-blocking InfiniBand Quantum at 800 Gb/s. Rail-optimized topology, native RDMA.
800 Gb/s
Layer 05
Orchestration
Sesterce OS - Slurm & Kubernetes native. Multi-tenant. Live DCGMI telemetry.
HIGH SLA
Newsroom
Latest announcements.
SoftBank Group and Sesterce to Develop 1 GW AI Data Center in Bosquel, France
Joint venture aims to advance France’s AI leadership, create jobs, and support AI adoption across Hauts-de-France.
Read release
Sesterce Appoints Brandon Records as SVP of AI Factories
Sesterce strengthens its leadership team to accelerate large-scale AI infrastructure deployment.
Read release
No Gradient Descent, Only Ascent by Sesterce, PyTorch, SemiAnalysis, GPU Mode: Inside the Paris Hackathon Pushing AI Systems Higher
Inside the Paris hackathon bringing Sesterce, SemiAnalysis, GPU Mode, PyTorch, Verda and Prime Intellect together to push AI systems performance higher.
Read releaseSesterce OS
One operating system.
Every AI Factory.
The control plane for sovereign compute. Provision clusters, orchestrate training, serve inference, and meter usage - across every Sesterce site, from a single pane of glass.
Zero-touch provisioning
From signed contract to live cluster in under 2 hours. No ticket queues.
Slurm + Kubernetes native
Run HPC training and production inference on the same fabric, same API.
Single-tenant by default
Dedicated hardware, dedicated network, dedicated storage. No neighbors.
Rack-level observability
Real-time telemetry at the rack, node, GPU, and job level - with APIs.
Customer stories
What customers say
about Sesterce.
Real conversations with technology and infrastructure partners building the sovereign AI stack with Sesterce.
1:48
Sebastien Urios
EMEA Sales Director
Flex AI
“
Sesterce gives us European capacity for critical AI workloads.
2:55
Dan Chester
EMEA Sales Director
VAST Data
“
Sesterce connects the infrastructure layers enterprise AI needs.
Sesterce CLI
Sesterce CLI.
Self-service access to every Sesterce AI Factory. Launch from a single command or reserve multi-megawatt pods - API-first, pay only for what you use.
API & CLI first
Launch, scale, and destroy clusters from a single command. Every action is an API call.
On-demand · reserved · spot
Three billing modes, same clusters. Pay-per-second for bursts, commit for frontier runs.
Serverless inference
Deploy Llama, Qwen, DeepSeek endpoints in 2 clicks. Sub-50 ms P95 latency.
Terraform
Official providers and operators. Drop-in to your existing IaC and GitOps pipelines.
B200New
$4.50 / gpu / hrH200
$2.25 / gpu / hrH100
$1.66 / gpu / hrA100
$1.24 / gpu / hrSesterce cloud
Best GPU rates & models
in 5 clicks.
Available through cloud.sesterce.com - live on-demand offers by GPU model, with availability and region coverage from Sesterce Cloud.
NVDA·B300 $7.91 ▲ liveNVDA·B200 $4.11 ▲ liveNVDA·H200 $2.09 ─ limitedNVDA·H100 $1.83 ▲ liveNVDA·A100 $1.36 ▲ liveNVDA·L40S $0.76 ▲ liveNVDA·B300 $7.91 ▲ liveNVDA·B200 $4.11 ▲ liveNVDA·H200 $2.09 ─ limitedNVDA·H100 $1.83 ▲ liveNVDA·A100 $1.36 ▲ liveNVDA·L40S $0.76 ▲ live
| Model | From / GPU / hr | Available now | Listed GPUs | Live regions | Action |
|---|---|---|---|---|---|
B300 NVIDIA Blackwell · SXM6 | $7.91 | 9 GPUs | 23 GPUs | 2/7 regions | Book → |
B200 NVIDIA Blackwell · SXM6 | $4.11 | 3 GPUs | 46 GPUs | 2/15 regions | Book → |
H200 NVIDIA Hopper · SXM5 | $2.09 | Limited | 74 GPUs | 0/30 regions | Book → |
H100 NVIDIA Hopper · SXM5 / PCIe | $1.83 | 22 GPUs | 152 GPUs | 14/85 regions | Book → |
A100 NVIDIA Ampere · SXM4 / PCIe | $1.36 | 25 GPUs | 199 GPUs | 13/148 regions | Book → |
L40S NVIDIA Ada · PCIe | $0.76 | 18 GPUs | 74 GPUs | 9/43 regions | Book → |
Reserve capacity
Tell us what you
need to build.
From a single GPU on-demand to a multi-megawatt training pod - share your workload and our team will respond within one business day.
1 business day response from a dedicated solutions architect
Custom commercial terms - on-demand, reserved or co-location
NDAs and DPAs ready to execute, sub-processor list public