Run AI models with enterprise-grade security without sacrificing performance. Phala Cloud Confidential AI protects your models and data using GPU TEE - hardware-isolated environments that keep your AI workloads private and verifiable.

​
Why Confidential AI?

Traditional cloud AI deployments expose your models and data to the cloud provider. Confidential AI solves this by running everything inside hardware-protected TEE. Your models stay private, your data stays secure, and you get cryptographic proof that execution happened in a trusted environment.

​
Try Confidential AI

Confidential AI API

Try LLM inference API nowPre-deployed LLM inference API with OpenAI-compatible interface

Confidential AI Models

Try inference your own AI modelsDeploy and manage AI models in a secure, confidential computing environment

Confidential GPU

Try to deploy your own models on GPU TEERent dedicated GPU TEE servers for custom model deployment

​
Quick Through from Confidential AI

​
API and Models

Here are two products, Confidential AI API and Confidential AI Models. Confidential AI API provides a pre-deployed LLM inference service with an OpenAI-compatible interface, making it easy to integrate into your applications. And Confidential AI Models enables you to deploy and manage your own AI models in GPU TEE.

​
Confidential GPU

If you want to deploy custom model with complete control over your infrastructure, check Confidential GPU to depoy with various GPU configurations, configure CPU, RAM, and storage to match your workload requirements based on your needs.

​
Benchmark

Our performance benchmark shows TEE mode on H100/H200 GPUs runs up to 99% efficiency, nearly matching native performance. This means you get confidential computing with minimal performance penalty.

​
FAQs

Check FAQs for frequently asked questions about Confidential AI.

​
What makes Phala Cloud Confidential AI Different?

API and Models

Pre-deployed LLM inference API with OpenAI-compatible interface

Confidential GPU

Rent dedicated GPU TEE servers for custom model deployment
  • Use existing code: Drop-in replacement for OpenAI APIs
  • Popular models: DeepSeek, Llama, GPT-OSS, and Qwen models ready to use
  • Verify execution: Get attestation reports proving your code ran in TEE
  • Pay as you go: Only pay for what you use
  • Custom models: Run your own fine-tuned or proprietary models
  • High performance: H200, H100, and B200 GPUs available
  • Full control: Configure CPU, RAM, storage, and location
  • Flexible scaling: 1 or 8 GPUs per server, various commitment options

​
Open Source Foundation

Our underlying technology is open source. Check out the private-ml-sdk repository to see how LLMs run securely in GPU TEEs. This project was built by Phala Network with support from NEARAI.

Open Source Implementation: Private ML SDK