# AI Servers Source: https://docs.infraista.com/en/compute/ai-servers An **AI server** is a dedicated GPU server: one or more NVIDIA GPUs, with the drivers pre-installed and full root (or Administrator) access. It works like an [instance](/en/compute/instances), with GPU plans and GPU metrics. | | AI Servers | Instances | | -------- | ------------------------------------------- | --------------------------------- | | Hardware | Dedicated NVIDIA GPUs plus CPU, RAM and SSD | Virtual CPU, RAM and SSD | | Good for | Rendering, VFX, training, inference | Apps, websites, APIs | | Connect | SSH, or Remote Desktop on Windows | SSH, or Remote Desktop on Windows | | Where | **Compute › AI Servers** | **Compute › Instances** |