Private AI servers
Run your team's AI assistant and models on a rack server you own. Prompts and documents stay on your network.
Runs vLLM · Open WebUI · Ollama
- From $34,999
- GPU memory 32–192 GB
- Ships in 7–10 business days
4 builds from $34,999 to $92,998
- 32GB GPU memory Threadripper PRO 7995WX AI Server with RTX 5090Threadripper PRO 7995WX · 256 GB ECC memory $34,999 View build
- 32GB GPU memory Threadripper PRO 9995WX AI Server with RTX 5090Threadripper PRO 9995WX · 512 GB ECC memory $52,999 View build
- 96GB GPU memory Threadripper PRO 9995WX AI Server with RTX PRO 6000Threadripper PRO 9995WX · 512 GB ECC memory $73,999 View build
- 192GB GPU memory Threadripper PRO 9995WX AI Server with Dual RTX PRO 6000Threadripper PRO 9995WX · 512 GB ECC memory $92,998 View build
Every server has
- Rack
- 6U rackmount · 17.2 × 10.4 × 20.6 in (W × H × D)
- Power
- SilverStone HELA 2500Rz · 2500 W 80 PLUS Platinum · ATX 3.1
- Network and remote access
- 2 × 10 Gb Ethernet (Intel X710) · IPMI (ASPEED AST2600) · 2 × USB4
- Cooling
- SilverStone XE360-TR5 360 mm liquid cooler
Which models fit
| Model | 32 GBRTX 5090From $34,999 | 96 GBRTX PRO 6000From $73,999 | 192 GB2× RTX PRO 6000From $92,998 |
|---|---|---|---|
| gpt-oss-20bOpenAI · 21B | Fits | Fits | Fits |
| Qwen3 32BAlibaba · 32B | 4-bit | Full | Full |
| Llama 3.3 70BMeta · 70B | Doesn't fit in GPU memory | 8-bit | Full |
| gpt-oss-120bOpenAI · 117B | Doesn't fit in GPU memory | Fits | Fits |
| Qwen3 235B-A22BAlibaba · 235B | Doesn't fit in GPU memory | Doesn't fit in GPU memory | 4-bit |
- Full16-bit, full quality
- FitsRuns as released
- 8-bitCompressed, close to full quality
- 4-bitCompressed, small quality drop
- Doesn't fit in GPU memory
Approximate GPU memory for the model plus a working context. Full means 16-bit. On two-GPU servers, vLLM splits one model across both cards.
Every build comes with
- Ready on first bootWindows 11 Pro or Ubuntu. Ollama, Open WebUI or ComfyUI set up on request.
- 48-hour burn-in report
- 3-year limited warranty
- Lifetime technical support
- Assembled in USA from new parts
- Built in 7–10 business days, shipped free
Questions
How many people can use it at once?
How is a server different from a workstation?
Which models can it run?
Does any data leave our network?
Does it come set up?
Your AI systems expert
- POs and purchase cards welcome
- Net terms
- University & government procurement
WHAT HAPPENS NEXT
-
An expert reads it.
Your models, budget, and preferences, reviewed by the person who would build the system.
-
You get a real spec.
Configuration with pricing you can hand to procurement, in one business day.
-
You decide.
No obligation. If a smaller build serves your models better, that is what we recommend.
