TURNKEY AI RACKS

Four racks. Four missions.

Racks are the products we deliver. Each rack combines several pre-validated solutions into one integrated system, built on our three hardware platforms. You order the rack; everything inside comes with it.

Request a Rack Quote Compare the racks →

Shipping DF-R1

Training Rack

Multi-GPU LLM and vision-model training clusters. Fabric-attached NVMe storage. Hybrid memory expansion for large working sets.

Explore Training Rack → Shipping DF-R2

Inference Rack

Production LLM serving at enterprise scale. Xeon-based inference nodes. NVMe model cache with sub-second cold-swap.

Explore Inference Rack → Shipping DF-R3

Edge Rack

Ruggedised half-rack for remote sites. On-board environmental sensing. Resilient WAN with dual-SIM cellular fallback.

Explore Edge Rack → Shipping DF-R4

Composable Rack

PCIe-any-to-any fabric. GPU, NVMe, and accelerator endpoints that any host can claim on demand. No vendor lock-in.

Explore Composable Rack →

SIDE BY SIDE

Choosing the right rack.

Same backplane, same DCIM, same delivery discipline. The differences are in the compute, storage, mechanical, and networking profile.

	DF-R1 Training	DF-R2 Inference	DF-R3 Edge	DF-R4 Composable
Primary mission	GPU training	LLM serving	Remote-site AI	Composable pools
Form factor	42U full rack	42U full rack	24U half-rack	42U full rack
Power envelope	~30 kW	~18 kW	~6–8 kW	~20 kW
Environmental	Datacenter	Datacenter	Industrial · sealed	Datacenter
Storage tier	TB/s to GPUs	Model cache	Local NVMe	JBOF pool
Memory expansion	✓ hybrid tier	Optional	—	Optional
WAN resilience	Standard	Standard	Dual-SIM + sat	Standard
DCIM modules included	All 8	All 8	All 8	All 8

Not sure which rack? We can help you pick.

Tell us the workload and the constraints. We will recommend a rack, a bill of materials, and a delivery date within one business day.

Request a Rack Quote