Infrastructure
L1 · ZB Compute
ZB Compute is the compute infrastructure that runs the models, the agents and your finished solution.
We are starting with 64 GPUs, the specialized processors used to power AI.
What ZB Compute runs
- Open models. Open models in the network run directly on ZB Compute, on GPUs sized to the work in front of them.
- Agents. Each agent works in its own sealed computer, with its own files and tools: code, browser, files and tests.
- The finished product. When a project launches, ZB Compute hosts it and keeps it running.
- The hardware protection that confidential computing uses.
One AI project needs two kinds of compute, model compute and agent compute, and ZB Compute runs both. ZB Model can coordinate open-model inference on ZB Compute with calls to frontier models through the providers that host them, so each task can use the model suited to the work.
You choose where your work runs
| Where | What it means |
|---|---|
| On any cloud | Compute on demand, in the regions you choose. |
| In your private cloud | ZB Compute inside your own cloud account. |
| Inside your border | Compute that stays in your country, which is the ground a sovereign solution stands on. |
Private and in-country deployments are scoped with each enterprise or nation. Bring the intended environment, the data that must stay there, and the models and agents your job uses.
Why it completes the stack
Owning this layer completes the stack. A person who starts with a conversation has the servers, the cloud and the deployment already taken care of.
How it connects to the stack
Receives approved actions from Security. Runs everything above it.