AI inference

Phala Cloud

Confidential AI cloud — run agents, LLMs, and GPU jobs inside hardware-backed TEEs with verifiable results

phala.com

Verified API available

Quick facts
What is it	Confidential AI cloud — run agents, LLMs, and GPU jobs inside hardware-backed TEEs with verifiable results
Pricing	Paid
Free tier	No
Platform	Web Application
API	Yes
Best for	Deploying AI agents with private execution and verifiable runtime proofs, Running sensitive LLM workloads in regulated industries
Domain registered	2008

Data updated Aug. 1, 2026

What does Phala Cloud do?

Phala Cloud is a confidential AI cloud platform that lets you run AI agents, private LLM models, and GPU jobs inside hardware-backed Trusted Execution Environments (TEEs). Instead of asking users to trust a cloud provider's claims, Phala Cloud emits runtime measurements that software can verify — so you know exactly what code ran and that your data stayed private. You deploy using familiar Docker Compose files, and the platform handles the rest: encrypting environment variables, selecting CPU or GPU TEEs (Intel TDX or NVIDIA CC), and spinning up confidential VMs.

The platform works through a CLI tool you install via npm. A single command — `phala deploy -c docker-compose.yml -n myapp` — builds a compose plan, selects the right TEE, injects encrypted environment variables, and boots your app. You can then query attestation data via an API endpoint to verify the runtime state. Phala Cloud also offers a GPU marketplace with H100, H200, and B300 capacity starting at $3.20 per GPU hour, plus OpenAI-compatible LLM endpoints for models from Z.ai, Qwen, DeepSeek, Google, and others. Every model endpoint is marked "encrypted" and runs inside a TEE.

This platform is built for developers and enterprises that need to run AI workloads with strong privacy guarantees — think regulated industries, financial services, healthcare, or any scenario where data confidentiality is non-negotiable. It's also useful for AI agent developers who want to deploy backends with sealed keys and verifiable execution. With over 5,000 users and partnerships with Nvidia, Intel, OpenRouter, and others, Phala Cloud is already proven at scale for production AI workloads.

AI API AI inference Developer Tools LLM

#ai infrastructure#attestation#confidential computing#docker-deployment#gpu-marketplace#llm hosting#privacy#tee

Key features

What makes it stand out

Deploy Docker Compose workloads into CPU or GPU confidential VMs with one command

Hardware-backed TEEs (Intel TDX, NVIDIA CC) keep code and data encrypted during execution

Attestation API emits runtime measurements that software can verify, proving what ran

OpenAI-compatible LLM endpoints with private prompts and verifiable runtime state

GPU marketplace with H100, H200, B300 capacity starting at $3.20/GPU/hr

Who is Phala Cloud for?

Who benefits most from this tool

Deploying AI agents with private execution and verifiable runtime proofs

Running sensitive LLM workloads in regulated industries

Hosting GPU-accelerated AI jobs with hardware-level confidentiality

Pricing

Usage

Custom

Confidential VM from $0.06/hour
GPU TEE from $3.80/hour
Storage $0.000139/GB/hour
tdx.small: $0.06/hour, 1 vCPU, 2 GB RAM, 20 GB disk
tdx.medium: $0.12/hour, 2 vCPU, 4 GB RAM, 20 GB disk
tdx.large: $0.23/hour, 4 vCPU, 8 GB RAM, 20 GB disk

Trust & presence

Domain Domain registered 2008

Gallery

Click any image to enlarge

Phala Cloud — pricing page — Pricing page

Alternatives in AI inference

cirrascale.com AI inference

Cloud platform providing on-demand access to multiple AI accelerators for development, training, and inference workloads.

Prem AI inference

Confidential AI stack for enterprises — run private AI workloads with hardware-verified encryption and zero data exposure

Akamai AI inference

Cloud infrastructure platform for deploying low-latency apps with GPUs, Kubernetes, and flat pricing

n8n Top 1k site

ThirdAI AI inference

Deploy private GenAI on CPUs — build chatbots, search, and AI agents in days without GPUs

Denvr AI Cloud AI inference

High-performance AI cloud platform for training, inference, and data science with sovereign data centers in Canada and USA

SiliconFlow AI inference

AI model inference platform — access multiple LLMs and multimodal models through a single API with predictable pricing

n8n

Chutes AI inference

Serverless AI compute platform for running open-source LLMs, image, video, and audio models at scale

n8n

TextSynth AI inference

Access large language, text-to-image, and speech models via REST API and playground

Similar tools

VELA Developer Tools

Open-source secure code execution runtime for AI agents — runs untrusted code in isolated Firecracker micro-VMs.

dstack Developer Tools

Open-source platform to provision GPUs and orchestrate AI workloads across clouds, Kubernetes, and on-prem clusters.

Orgn Developer Tools

Confidential AI stack for regulated industries — secure IDE, model gateway, and agent orchestration with verifiable audit trails

Kern AI refinery AI Agent

Enterprise platform for building secure, confidential AI assistants and knowledge agents with data privacy and compliance.

HiddenLayer Pentesting

Security platform that protects AI models and applications from adversarial attacks, supply chain risks, and model theft.

Openlayer Testing

AI governance platform that monitors, tests, and secures your AI systems from development to production.

Cognitora Developer Tools

Cloud platform providing autonomous AI agents with secure sandbox environments for code execution and data analysis

Palapa Developer Tools

Share your desktop GPU as a cloud endpoint for AI apps, accessible from anywhere via an OpenAI-compatible API.