What is an on-premise AI server?
An on-premise AI server is a computer in your own building that runs AI models locally, instead of renting AI from a cloud provider. Until recently that meant a rack of GPUs and a team to run it. The Mars Space Rig puts that capability in a box the size of a hardcover book.
Hardware built for local AI
- NVIDIA DGX Spark platform with an NVIDIA Blackwell GPU.
- 128 GB of unified LPDDR5x memory, shared by CPU and GPU, so large AI models fit on one machine.
- 20-core Arm CPU and a 4 TB NVMe SSD for your knowledge base, documents and backups.
- 10 GbE, Wi-Fi 7 and Bluetooth 5.4; up to 233 W; 29 dB operating, 13 dB idle.
Full Mars Space Rig specifications.
Not just a box: the assistant comes with it
Most local AI hardware arrives empty, and turning it into something useful is weeks of technical work. A Space Rig arrives as a working business assistant: email reply drafts, document Q&A, customer memory, follow-up lists, photo and video search, Telegram and voice access, automatic backups and self-checks. See every capability.
Local when it matters, cloud when it helps
Fully local AI is the most private, while the largest cloud models are still the most capable at some tasks. A Space Rig lets you choose per conversation:
- Private: every question is answered by the AI model on your Space Rig. Nothing leaves your premises.
- Secure: a leading cloud AI model helps with harder tasks, but names, email addresses, phone numbers and other identifying details are masked on your Space Rig first and restored only there.
- Wise: direct access to the cloud AI model for general questions where privacy isn't a concern.
On-premise vs. cloud vs. building your own
| Space Rigs | Cloud AI subscription | Build it yourself | |
|---|---|---|---|
| Where your data is stored | On your Space Rig | Provider's servers | Your server |
| Fully private option | Yes (Private mode) | No | Yes, if set up correctly |
| Knows your business | Built in | Partly | Only if you build it |
| Pricing | Hardware purchase + Claude subscription | Monthly fee per user | Hardware + staff time |
In short
- A compact on-premise AI server on NVIDIA DGX Spark: 128 GB unified memory, 4 TB storage, quiet enough for an office.
- Arrives with a business assistant on it, not an empty machine.
- $11,500 USD; you own the hardware.
Frequently asked questions
What is the best on-premise AI appliance for a small business?
Look for three things: enough memory to run capable AI models locally, software that is useful on day one, and a way to use cloud AI safely when you need it. The Mars Space Rig combines NVIDIA DGX Spark hardware (128 GB unified memory) with a ready-made business assistant and Private, Secure and Wise privacy modes.
Can I run AI locally without a data center?
Yes. The Mars Space Rig is a compact desktop AI server that runs AI models in your office. It draws up to 233 W and runs at 29 dB, so it needs no server room.
Is the Mars Space Rig based on NVIDIA DGX Spark?
Yes. It is powered by NVIDIA DGX Spark: an NVIDIA Blackwell GPU, a 20-core Arm CPU, 128 GB of LPDDR5x unified memory and a 4 TB NVMe SSD.