Autonomous Computer 2

From

£19,627.20

Computer

Autonomous Computer 2

BuildBuild: Complete machine · 2× RTX 5090

Complete machine · 2× RTX 5090

Complete machine · 2× RTX PRO 6000

DIY kit · Case and PCIe 5.0 risers

Operating system

Omarchy

Ubuntu

Memory

64 GB

128 GB

192 GB

256 GB

Storage

1 TB

2 TB

4 TB

8 TB

Buy more, save more

Qty	Price
3-4	£19,039
5-9	£18,646
10-29	£18,449
30+	£18,254

£19,627.20

Free shipping within 4 weeks

Open source

Autonomous Computer 2

Free shipping within 4 weeks

£19,627.20

Update #3

Fresh of the mill.

DeeCreatorOct 6, 2026

Another 2x RTX PRO 6000 Blackwell case's out of the machine this morning.

Every chassis gets test fitted before anodizing, so nothing surprises us after the finish is on.

Teaser clip for the update on being cut off by model providers

Update #2

You are one step away from being blocked by OpenAI or Anthropic any time.

DeeCreatorAugust 29, 2026

In June, the US government took Fable 5 from you. Today, OpenAI blocks you for using Cursor. Who cuts you off next?

Build your own computer.

Update #1

This is what 64GB of VRAM on a desk does.

DeeCreatorAugust 23, 2026

Our Autonomous Computer 2x RTX 5090 built this before lunch. One prompt into OpenCode, Qwen3.8-27B did the rest, one HTML file out.

Own your intelligence.

Never revoked.

Your models don't get deprecated, rate-limited, or pulled. The weights sit on your disk.

Private by physics.

Prompts never leave the building. Nobody trains on your work. Not by policy, by physics.

No per-token bill.

Own the machine and agents can run around the clock. You pay for electricity, not tokens.

Yours to keep.

Fine-tune on what only you know. The model becomes an asset you own, and it sells with the company.

GPUs

Two cards, tuned and burn-tested. We install RTX 5090 or RTX PRO 6000 Blackwell— installed, driver-loaded, and benchmarked as one system before it ships. Plug it in and run.

64–192GB

Total VRAM

210–252TFLOPS

FP32

3,584GB/s

Memory bandwidth

Two RTX 5090 cards mounted vertically in the open chassis frame

Someone holding the finished machine, triangular cutouts across the panel

Close-up of the top corner, a GPU fan behind the vent panel

Chassis

CNC-machined aluminum chassis. Not a bent-steel box — precision-cut and milled from a solid 5mm aluminum plate. It moves heat like a heatsink; the triangular cutouts keep it stiff without dead weight. 12.5 × 12.5 × 16 inch, anodized black.

33lb

Weight

CNC

Solid aluminum

Close-up of the milled triangular cutouts in the aluminium panel

Machined aluminium panels laid out before assembly

Riser

A full ×16 to every card. Each GPU reaches the board over its own server-grade PCIe 5.0 riser — shielded, custom-cut, no dropped lanes. Every power run is cut to length and sleeved by hand.

x16

Lanes per card

64GB/s

Per card

Gen 5

PCIe

The open frame from the side, a GPU on its riser above the board

Both GPUs seated on their risers with sleeved power runs

Motherboard

ASUS Pro WS W790-ACE. The hardest choice in a multi-GPU build, solved with a workstation board: every card gets a full ×16, and the BIOS ships pre-tuned so every card links at full width.

PCIe 5.0 ×16 slots

DDR5 channels

W790

Chipset

The workstation board with CPU coolers, memory and PCIe slots

Close-up of the board's PCIe slots beside the heatsink

A gloved hand lowering the board into the chassis

CPU

Intel Xeon w5-3425. The lanes are the point: 112 lanes of PCIe 5.0 means both cards, the NVMe, and the risers never fight for bandwidth.

112

PCIe 5.0 lanes

Cores

4677

Socket

Power

2000W that doesn't flinch. Under full load the box pulls up to 1600W — and cheap power is where multi-GPU builds die. We spec 2000W with native 12V-2x6, feeding clean power to every card. The result is boring: it just runs.

2000W

Power supply

1600W

Max draw

The open machine from the side, power supply mounted low

A gloved hand fitting the power supply into the chassis

The power supply's modular sockets with cables routed to the board

Cooling

Air moved on purpose. Six fans on a single controller, mapped to the airflow path before a panel was cut — not bolted on after.

Two fan trays lifted out, wiring looped between them

Close-up of a fan tray being placed into the chassis

Fitting a fan panel beside the mounted GPUs

Memory & Storage

ECC memory, fast NVMe. Error-correcting memory catches the silent bit-flip before it corrupts the job you left running overnight. The NVMe loads a 70B checkpoint in seconds, not minutes. Need more? Configure up to 256GB RAM and 8TB Gen5 NVMe.

128GB

ECC RAM

2TB

NVMe

7GB/s

Read

Four ECC memory sticks laid out in a row

ECC sticks installed beside the CPU heatsinks

Inside the built machine, memory and cabling behind the panel

Performance

Concurrent usersPer-user generationTotal tok/sTTFT (s)

4 users76.8Total tok/s840TTFT (s)0.64 s

8 users72.1Total tok/s1,488TTFT (s)1.15 s

16 users64Total tok/s2,390TTFT (s)2.27 s

24 users52Total tok/s3,004TTFT (s)2.41 s

How big is your team?

vLLM • FP8 weights + KV • 1K prompt • 512 output

Qwen3.6-27B, per user (tok/s)

Grey: Mac Studio M3 Ultra single user, published run. Blue: this box on vLLM FP8, per user while serving 4 and 24 users at once.

Mac Studio M3 Ultra 512GB, 1 user (MLX 4-bit)

Mac Studio M3 Ultra, 1 user, best tuned build (MTPLX v2)

2x RTX 5090, EACH of 4 users (this box)

2x RTX 5090, EACH of 24 users (this box)

Source:Community benchmark

What Qwen3.6-27B gets done in one hour.

Runs these models

Language6

Qwen3.6 27B
Qwen3.6 35B A3B
Gemma-4 31B
Gemma-4 26B A4B
Laguna-S 2.14-bit GGUF
Laguna-XS 2.1

Image3

Qwen Image 2512
Qwen Image Editing 2511
FLUX.2-dev

Video2

Wan 2.2
MiniMax H3

Quantized builds noted per model. What fits depends on context length and KV cache.

GPU

Cards

2x NVIDIA RTX 5090 or 2x NVIDIA RTX PRO 6000 Blackwell

Total VRAM

64 GB (RTX 5090) / 192 GB (RTX PRO 6000)

FP32 compute

210 TFLOPS (RTX 5090) / 252 TFLOPS (RTX PRO 6000)

Interconnect

PCIe 5.0 x16 per card, Gen5 riser cables

Compute

CPU

Intel Xeon w5-3425 - 12 cores, 112 PCIe 5.0 lanes

Motherboard

ASUS Pro WS W790-ACE - 5x PCIe 5.0 x16

CPU cooler

Heatsink Dynatron S6

Memory & storage

RAM

128 GB (4x 32GB DDR5-4800 ECC) - configurable 64-256 GB

Storage

2TB NVMe PCIe 4.0 - configurable 1-8TB, Gen5 on 8TB

Power & cooling

Power supply

2000W, native 12V-2x6

Fans

6x Thermalright TL-N12-R9, single controller

Dimensions

Size

13.9"L x 13.9"W x 18.2"H

Weight

33 lbs

Chassis

CNC-machined solid aluminum, anodized black

Software

Ubuntu, NVIDIA drivers pre-installed

AI frameworks

PyTorch, CUDA, vLLM, SGLang, llama.cpp

Warranty

Full system

1 year

Other

Open source

CAD, BOM & BIOS on GitHub

Shipping

4 weeks

Dimensions diagram of the machine, 13.9 by 13.9 by 18.2 inches

The lineup

Three builds.

Autonomous Computer (2-GPU)

13.9” × 13.9” × 18.2” - 33lb

RTX Pro 6000 Blackwell or RTX 5090 x2
Up to 192GB VRAM - 128GB ECC RAM
Intel Xeon w5-3425 - 2TB NVMe
CNC-machined aluminum, black
PCIe 5.0 riser x2 - 6-fan cooling
2000W power

Autonomous Computer Studio (4-GPU)

19.8” × 19.8” × 24.2” · 66lb

RTX Pro 6000 Blackwell or RTX 5090 x4
Up to 384GB VRAM - 192GB ECC RAM
Intel Xeon w5-3425 - 2TB NVMe
CNC-machined aluminum, silver
PCIe 5.0 riser x4 - 18-fan cooling
2x 2000W power - 240V circuit

Autonomous Computer Rack (8-GPU)

17.3” × 35.5” × 12.2” - 93lb

RTX Pro 6000 Blackwell or RTX 5090 x8
Up to 768GB VRAM - 512GB ECC RAM
2x AMD EPYC 9124 - 8TB NVMe hot-swap
Steel 7U chassis, hot-swap everything
8x PCIe 5.0 x16 via MCIO - 12 hot-swap fans
2000W CRPS redundant - 240V circuit

Open-source hardware

The whole machine
is open.

Every CAD file, bill of materials, and BIOS setting — free on GitHub. Fork it, change it, build your own, even sell it. This is how the personal computer began: in the open. We’re doing it again, for AI hardware.

GitHub 1.5k

autonomous-ai / autonomous-computer

2x-5090/ 4x-5090/ 8x-5090/ 4x-6000/

bom/ step_models/ stl-models/ photos/

README.md setup.md

MIT · 175 forks

Clone it. Build it. Sell it. We just want it built.

Autonomous Computer 2