01
Yours, not shared
Single-tenant nodes with their own compute, storage and network. No noisy neighbours and no spot-market eviction.
Bratislava · Live
The network
One site live. The model repeats.
Bratislava sits inside the EU, an hour from Vienna, with lower operating costs and strong technical talent in Central Europe. The next sites follow the same model: leased data-centre space, Sapience-operated Blackwell clusters, EU jurisdiction. Locations are announced when they are contracted, not before.
Bratislava, SlovakiaLive · 11 MW identified · 5 MW deploying
First HGX B300 clusterDeliveries from December 2026
Next sitesFurther EU locations under evaluation; announced when contracted
JurisdictionSlovakia, European Union
How we build
Dedicated hardware, delivered on a date.
We do not oversubscribe, and we do not build data centres: we lease space and power in existing data centres so capital goes into GPUs. Each cluster is dedicated to one customer, commissioned on an agreed delivery date, and run by our own team. The model repeats site by site.
02
Blackwell only
Every node is current-generation NVIDIA Blackwell in 8-GPU systems built by Supermicro: HGX B300 for training-scale clusters, RTX PRO 6000 Server Edition for inference and agentic workloads.
03
Operated end to end
We procure the hardware, run the fabric and operate the stack, with 24/7 monitoring by our own engineers. No hyperscaler layer between you and the GPU.
Built with
Need a private cluster with data residency?
Talk to us about isolated environments, contractual data residency and delivery dates.