Twenty systems. One configuration.
We chose one machine and are building the fleet out of it twenty times. A single, well-understood configuration is easier to operate and support than a catalogue, and every tenant gets the same hardware.
GPU8 × NVIDIA HGX B300
Host2 × x86 CPU · 24 × DDR5-6400
Network8 × 800GbE / XDR800
| GPU | 8 × NVIDIA HGX B300 per system, 160 in total |
|---|---|
| Host | dual-socket x86, 1.5–2 TB DDR5-6400 |
| Network | 8 × 800GbE / XDR800 per system |
| Power | approx. 14.7 kW per system, around 300 kW across the fleet |
| Cooling | air-cooled throughout |
| Site | carrier-neutral colocation, United Arab Emirates |
| First capacity | Q4 2026 |
Data residency
The hardware is ours and it is installed in the United Arab Emirates. Compute runs in the country and data stays in the country.
Renting
Directly from us on contract, or through the marketplaces — our capacity will be listed on Vast.ai and RunPod. Pricing on enquiry.