Pick how much power you actually need.
The same software at every size: from a box for one office to a server for a whole plant. It installs on your premises and works inside your network, with no usage-based subscription.
Pick your configuration. Five levels of power.
It starts with a box for one office and ends with a server for a whole plant, always on NVIDIA hardware. Scale when you need to, with nothing to redo.
Edge
Jetson Orin Nano
AI compute
67 TOPS
Memory
8 GB
Ideal for: Prototypes and compact sites
Advanced edge
Jetson AGX Orin
AI compute
275 TOPS
Memory
64 GB
Ideal for: Departments, branches and offices
Edge AI
Jetson Thor
AI compute
2070 TFLOPS
Memory
128 GB
Ideal for: Heavy use, all on site
Desktop AI
GB10 Grace Blackwell
AI compute
1 PFLOP
Memory
128 GB unified
Ideal for: A whole company, even the largest models
Datacenter
Custom server
AI compute
Multi-GPU
Memory
Tailored
Ideal for: Clusters, high capacity and redundancy
Figures from NVIDIA public specifications; the final configuration can vary from model to model.
Change the hardware, not the experience.
Assistants, documents and access permissions are the same from the smallest machine to the server.
Same software
Assistants, document templates and connections to your apps: identical on every Box.
Encrypted storage
NVMe Gen4 up to 8 TB with AES-256 encryption of data at rest.
Full connectivity
High-speed wired and wireless networking, plus Thunderbolt expansion.
Runs inside your network
The AI and your documents stay on the Box, on your premises. Cloud connections are optional and always declared.
Cluster ready
Add more boxes and they collaborate as a single compute pool.
No maintenance
It sets itself up and updates itself: no technical skills required.
Storage
up to 8 TB NVMe
Encryption
AES-256
Network
10GbE RJ45
Wireless
Wi-Fi 6E
I/O
Thunderbolt 4
Let’s start with your work. See what AI can do.
Tell us about a task your team handles. We’ll show you how to approach it with IntelligenceBox, from documents to the tools you use every day.