Castellan · Services · HPC Hosting
Service 07 — HPC Hosting & Colocation

Your hardware. Our racks, power, and uptime.

Castellan hosts GPU and HPC systems — from single HGX servers to rack-scale systems — in data centres in Europe and North America. High-density racks, redundant power, cooling engineered for accelerated hardware, and hands-on operations around the clock. You keep ownership of the machines; we keep them running.

2regions
Data-centre locations in Europe and North America
Redundant
Redundant power feeds, metered per rack
24/7
Monitoring, operations & remote hands
DLC-ready
Air & direct-liquid cooling for H100 → GB300
01
What's included

GPU-grade colocation, not generic rack space.

Accelerated hardware breaks the assumptions of classic colocation: an HGX H100 server draws more than many full legacy racks, and NVL72 systems require facility-level liquid cooling. Castellan places your systems in facilities engineered for exactly this class of hardware — and operates them as if they were our own.

  • 01
    Rack space & power
    High-density, liquid-cooling-ready racks with redundant, metered power feeds — from legacy systems to rack-scale deployments.
  • 02
    Cooling for accelerators
    Air, rear-door heat exchangers, and direct-liquid cooling — matched to your systems, from H100/H200 air-cooled servers to GB200/GB300 liquid-cooled racks.
  • 03
    Network & interconnect
    Redundant transit, private interconnects, and low-latency cross-connects — plus InfiniBand-ready cabling inside your deployment.
  • 04
    Remote hands & lifecycle
    Installation, burn-in, component swaps, firmware and lifecycle management — 24/7, by engineers who work with this hardware daily.
02
Why host with Castellan

Six reasons owners bring their systems to us.

01
Transparent power costs
Metered power, passed through transparently with monthly reporting — on hardware where electricity is the dominant cost of ownership.
02
GPU-grade facilities
Densities, cooling, and floor loading specified for accelerated hardware — no retrofitting, no compromises.
03
Jurisdiction and reach
Locations in Europe under Swiss and EU law with GDPR/revFADP-aligned operations — and in North America where you need it.
04
One team for onboarding
From signed order to energised systems — logistics, staging, installation and acceptance handled by one team.
05
Monetize idle time
Hosted systems can join our Compute Monetization marketplace: contracted buyers take your idle GPU-hours, turning cost into revenue.
06
A path to scale
Grow from hosted racks to a full data centre leased from Castellan — see Powered Shell & Turnkey.
03
Engagement

Clear terms, professional operations.

Hosting engagements range from a single system to multi-rack deployments, with transparent power pass-through and defined service levels. Combined with GPU Procurement & Trading (we source and deliver the hardware) and Compute Monetization (we bring buyers for your idle capacity), we cover the full lifecycle of owning AI hardware.

Scale
Single servers · rack deployments · multi-rack clusters
Supported systems
NVIDIA H100 · H200 · B200 HGX servers · rack-scale GB200/GB300 NVL72, site-dependent · storage & networking
Locations
Data centres in Europe and North America · further locations on request
Power & billing
Metered power pass-through · transparent monthly reporting
Operations
24/7 monitoring · remote hands · SLA-backed response times
Options
Idle-time monetization · hardware procurement & refresh · managed operations
04
Hosting enquiry

Tell us about your systems.

Describe your hardware, preferred location, and timeline — we come back with available capacity, power pricing, and an onboarding plan. Prefer email? info@castellan-intelligence.com