Our builds

Servers built for the models you run.

Tell us the work. We build the machine.

We spec, build and test AI servers for teams that want their models on their own hardware.

Request a buildSee the current build
  • Built around your models
  • Tested with Bridge
  • Yours to keep
  • Rack or tower
A workstation motherboard from one of our machines, seen from above, with an NVIDIA graphics card, a liquid-cooled processor and eight memory sticks
One of our own machines.

On the bench now

A rack server with several RTX PRO 6000 cards.

A recent build: a 6U rack server for large models and many people at once.

Build sheet

GPUs
NVIDIA RTX PRO 6000 Blackwell, several cards96 GB of memory on each card
Processor
AMD Ryzen Threadripper PRO 9975WX32 cores, 64 threads
Motherboard
ASUS Pro WS WRX90E-SAGE SESeven PCIe 5.0 slots for GPUs, 10 Gb networking
Memory
128 GB DDR5 ECC registeredEight channels, 8 × 16 GB
Storage
Two Samsung 9100 PRO 4 TBPCIe 5.0 NVMe, for model files
Cooling
SilverStone XE360-TR5360 mm liquid cooler for the processor
Chassis
SilverStone 6U rackmountOn ball-bearing rails
Power
Sized for every card at full loadEach RTX PRO 6000 can draw up to 600 W

What it can run

  • gpt-oss-120bAbout 65 GB: fits on one card
  • qwen3:235bTwo cards
  • Many people at onceRoom for every conversation on top of the model

Sizes from Models & hardware. Every conversation needs memory on top of the model, so more people need more.

How a build works

From the work to the machine.

  1. Tell us the work

    The models you want, how many people will use them, and what data they’ll handle.

  2. We spec it

    GPUs, memory, storage, power and cooling sized for that work, in a rack or a tower.

  3. We build and test it

    Assembled and tested with the models you’ll run, in Bridge.

  4. It goes to work

    On your network, with Bridge on every desk that needs it.

Request a build

Tell us what you want to run.

We’ll read it and reply by email with questions and a plan.

The build

Models, apps and jobs. For example: chat and coding for 20 people, image generation, or one large model.

Free rack units, power circuits and cooling, if you know them.

Your setup

Servers, GPUs or workstations you already own.

Patient records, contracts, source code and so on. No confidential details.

File shares, email, the apps your team uses.

Your IT team, an outside provider, or no one yet.

About you

Only if you have one.

We'll read this and reply by email. Don't include passwords or confidential data.

A few details

Questions, answered.

More in the full FAQ.

Who owns the server?

You do. It’s built for you and runs on your network. If you’d rather not run hardware, ask about Alluqi Managed: dedicated hardware Alluqi provisions and operates for you.

Can we rent a build?

Hardware rental is not offered today. Arrangements are case by case through sales.

What does Bridge need from the server?

A model engine, like Ollama or vLLM, on your network. Each person runs Bridge on their own computer and connects to it.