The full range, by class of work. Specifications are taken from the current product pages; where a figure is not verified it is not shown. Configuration options live on each system's own page.

Workstation-class machines for developing and running models on your own hardware, without a queue or a per-hour rate. One, two or four GPUs.

1× GPU

Epoch One

Local AI development, inference and fine-tuning

$8,450 Starting configuration

GPU	RTX 5090 / RTX 5000 ADA / RTX 6000 ADA
CPU	AMD Ryzen 9 9950X · 16-core
Storage	1TB or 2TB M.2 NVMe
Power	1000W PSU

Configure this system Chat with an engineer

2× GPU

Epoch Dual

Fine-tuning · multi-model inference

$21,800 Starting configuration

GPU	2× RTX 5090 / 2× RTX 5000 ADA / 2× RTX 6000 ADA
CPU	AMD Ryzen Threadripper PRO 7975WX · 32-core
Memory	128GB / 256GB / 512GB RDIMM
Storage	2TB M.2 NVMe · optional 2TB or 4TB SATA SSD
Cooling	AIO liquid

Configure this system Chat with an engineer

High density

Epoch Quad

High-density multi-GPU compute

Four-GPU compute, specified with an engineer against the workload, power and cooling it has to run in. Configured to order rather than from a price list.

Talk to an engineer

Rack-mounted compute for teams running private AI infrastructure rather than renting it. These are specified with an engineer against your rack, power and cooling constraints. There is no catalogue configuration to quote from, so nothing is listed here.

When a catalogue configuration is not the right answer. Tell us the workload, the GPU and memory it needs, the storage and deployment environment, and any form-factor constraint, and we will work out what the machine should be.

Single-GPU systems for content creation, CAD and simulation work, and for high-end gaming. Same assembly and testing process as every system we ship.