Server series

Lancer

Maximum payload for AI training.

Eight-GPU NVIDIA HGX B300 or AMD Instinct MI355X systems, air-cooled or liquid-cooled. Built for training and serving the largest models, and designed with your data-centre team.

Overview

Lancer is Allied's 8-GPU training platform: NVIDIA HGX B300 or AMD Instinct MI355X, air-cooled in 8U or liquid-cooled in 4U. Every Lancer is designed with you, integrated and burned in by Allied, and delivered ready to join your cluster.

Ideal for

Training and fine-tuning large language and multimodal models
Serving large models at high throughput
Research clusters and national AI programmes scaling from one node to many

What Allied delivers

Cluster design: GPU count, InfiniBand or Ethernet fabric, storage and power plan
Rack integration, cabling, burn-in and acceptance testing before shipping
On-site installation, software stack setup and ongoing support

Site requirements

Air-cooled models: 200–240 V circuits for full power-supply output (C19 cords); inlet air 10–30°C
Liquid-cooled models: a coolant distribution unit (CDU) and facility water loop; coolant inlet no higher than 45°C and above the dew point when humidity is over 50%