Hoster Overview
How to connect your GPU server to Qudata.ai and start renting out capacity.
Qudata Agent is installed on your server and turns it into a Qudata.ai marketplace node. After installation, the platform handles task execution automatically - you only need to keep the node available.
How it works
Each user task runs inside an isolated virtual machine.
Virtualization uses QEMU/KVM, and GPU is passed into VM using hardware PCI passthrough via VFIO.
Users get near-native hardware performance, while hosters cannot access task content.
You do not need NVIDIA drivers on host OS - drivers are installed automatically only inside VM.
How to become a hoster
- Prepare the server and enable IOMMU in BIOS/UEFI.
- Create an API key in the Developer section.
- Run the installer and follow the on-screen instructions.
- Return to My GPUs and submit the GPU for verification.
- After approval, set a price and publish the GPU on the marketplace.
Server requirements
| Parameter | Requirement |
|---|---|
| OS | Ubuntu 22.04+ |
| GPU | NVIDIA (without host drivers) |
| BIOS | VT-d / AMD-Vi enabled |
| RAM | from 8 GB |
| Disk | from 50 GB free |
| Access | root |
| Network | stable internet access without blocked outbound connections |
Installation flow
Run the command on the server, replacing the placeholder with your API key:
wget -qO- https://agent.qudata.ai/install.sh | sudo bash -s -- ak-YOUR-API-KEYInstallation runs in two stages and may require a server reboot.
Stage 1 - Host preparation
- Detect GPU and verify compatibility.
- Verify no NVIDIA drivers on host.
- Configure GPU isolation (IOMMU/VFIO) and install required components.
- Reboot if required. Installation resumes automatically afterward.
Stage 2 - Launch
- Download VM image.
- Run GPU passthrough test to verify transfer into VM.
- Start Agent as a system service.
After successful installation, the GPU appears in My GPUs with the Inactive status. Submit it for verification, wait for the Ready for Rent status, set a price, and publish it on the marketplace.