05 · Business model

Two product lines, launched in parallel

Model A secures base load and cash flow through long leases. Model B uses remaining capacity for higher revenue density. Both share the same power and data hall assets.

Product lines

Colocation and data hosting

Model A · Colocation

Base load

Rack and data hall rental
Customers bring their own GPU or HPC equipment
We provide power, liquid cooling, connectivity and staff
Pricing per kW, rack or dedicated data hall
Lower capital needs, long contracts, stable cash flow

Model B · Data hosting

Higher density

Owned compute and hosting services
We own GPU clusters for customer use
Bare metal, dedicated clusters and inference hosting
Higher revenue density per megawatt
Shared scheduling across sites creates a compute pool

Who we serve

Who uses our capacity

Enterprise AI
Cloud / Neocloud
AI and research
Hyperscaler overflow
Public sector

Data hosting architecture

Owned compute: grid connection to GPU racks

Planned 2–10 MW per site, sized to the load. PUE is a design target, verified after commissioning.

Power

MV supply · A/B distribution · UPS and backup power

Supply

Cooling

Liquid cooling and heat exchange

Target PUE 1.15–1.20

Racks

40–130 kW per rack · Liquid cooling · Circuit metering

Dense racks

Compute

256 GPUs per pod · Planned 4–16 pods per site

Computing

Network

Spine-leaf · 800G InfiniBand

Low latency

Storage

Parallel file system · NVMe hot tier · Object storage

Data

Management

Separate BMC network · DCIM · Security zones

Operations

Efficiency basis: EnEfG § 11. Waste heat recovery and renewable power are part of site planning.

Network planes

Four separate network planes

800G IB

Compute · East-West

Low-latency in-site network.

200G RoCEv2

Storage · Separate plane

Lossless Ethernet.

2 × 100G

Services · North-South

Multi-carrier BGP.

Out-of-band management

Dedicated physical links.

Site interconnection: Dark fibre · Latency validated by distance and use case
Revenue: GPU hours and monthly plans · Dedicated clusters · Inference hosting and APIs · Shared scheduling · Operations with SLAs

Develop · Own · Operate

Enquire about capacity

Colocation, dedicated clusters or inference hosting — talk to us.

Get in touch