Skip to main content

​
Overview

If the Confidential AI API and Model Template do not meet your needs, you can use GPU TEE to rent dedicated GPU servers with TEE protection for custom model deployment, inference, or fine-tuning. Perfect for confidential AI training and fine-tuning workloads on proprietary datasets. GPU TEE provides you GPU hardware with H100/H200/B200 series, NVIDIA Driver 570.133.20 and CUDA 12.8 installed. Learn more about NVIDIA H100 TEE, H200 TEE, and B200 TEE capabilities and specifications. Go to the GPU TEE tab in your Phala Cloud dashboard to get started.

​
Available hardware specifications

Firstly, choose GPU types from these TEE-enabled GPUs.

GPU Device Selection

Customize your setup with:
  • Cores and RAM: CPU cores and RAM allocation
  • Storage: SSD storage for datasets and models
  • Location: Choose from available data center regions
  • Scaling: Deploy 1, or up to 8 GPUs per instance
Once you select the GPU type, you can choose the GPU Count, usually between 1 to 8 GPUs, depending on your workload needs. Next, choose commitment period option from options, usually from 6 months, 3 months, 1 month, or on-demand based on your project duration.

​
Review & submit order

Check the Price Summary section to confirm your hardware specs and cost and Click Submit Order.

Order Review

After submitting, our team will contact you soon to confirm your setup and provide access credentials.