Tell us the models and the users. We return a configuration sheet and a quote.
Spec a system →Model serving, access control and monitoring on hardware you own.
See the platform →
Thirty minutes with an engineer: your workflow, your data, and whether custom AI fits.
Book a scoping call →
Four sizes of AI system, from a desk-side workstation to a multi-node cluster. We size each one from the models you'll run and the people who'll use them, source the parts across vendors, burn it in, and install it on your site.
Configurations are sized, not sold off a shelf. The ranges below are typical; the quote names every part and the reason for it.
Workstation-class, 1–2 GPUs, for a team or a pilot
| GPUs | 1–2 |
|---|---|
| GPU memory | 24–96 GB each |
| Users | ≈ 1–25 |
4–8 GPU rack server for production workloads
| GPUs | 4–8 |
|---|---|
| GPU memory | 48–192 GB each |
| Users | ≈ 25–500+ |
Multi-node, for large models and many users
| GPUs | 16 and up |
|---|---|
| GPU memory | 80–192 GB each |
| Users | Hundreds–thousands |
Compact or ruggedized, for offline sites
| Accelerators | Module or 1–2 GPUs |
|---|---|
| Memory | 16–64 GB |
| Users | One site |
We don't resell one manufacturer's line. For each system we compare GPUs, server platforms, networking and storage from several vendors on workload fit, availability, power and support, then write down the reasoning in a configuration sheet you keep.
That matters twice: once when you buy, because the parts suit the job, and again in three years, because nothing in the system ties your next purchase to us or to one vendor. How we source hardware explains the criteria, including how we choose GPUs.
Hardware is only half delivered when it leaves the factory. These four services come with every system.
Every system is assembled and run under sustained GPU, memory and storage load on our bench before it ships, with the cstmAI software loaded and your first use case running. Parts that fail early fail here, not in your server room.
We deliver, rack, cable and power the system, join it to your network and identity provider, and sign it off against an acceptance checklist with your IT team. Site surveys come first, so power and cooling are settled before delivery day.
GPU generations move quickly. We plan expansion and replacement at the start, track how loaded the system is, and recommend when adding a node or refreshing GPUs pays off, and when it doesn't.
A system built from several manufacturers' parts should still have one phone number. We register warranties, keep the records, and coordinate repairs and replacements with each manufacturer on your behalf.
Because the right configuration depends on your models, users and site, and GPU pricing moves month to month. Every system is quoted from a written configuration sheet, so you can see exactly what you're paying for.
Often, yes. We assess existing servers and GPUs in discovery and tell you plainly whether they fit the workload, need additions, or should be replaced.
Every system we deliver ships with the cstmAI software installed, because that is what we test and support end to end. Talk to us if you have a different arrangement in mind.
Tell us the models you want to run, how many people will use them and where the hardware should live. An engineer replies with a first configuration and the questions that decide the quote.
