Run Powerful AI GPUs,
With Security
Run NVIDIA H200/H100 workloads with complete hardware protection. The only cloud combining Intel TDX + NVIDIA Confidential Computing for end-to-end security.
Available GPU Configurations
From single GPU to enterprise clusters. All configurations include Intel TDX + NVIDIA Confidential Computing protection.
NVIDIA H200
NVIDIA H100
NVIDIA B200
All prices include Intel TDX + NVIDIA Confidential Computing protection. Volume discounts available. Contact sales for custom configurations.
Private GPU like Local,
But No DevOps Troubles
Most clouds offer GPU-only protection or no TEE at all. Phala Cloud protects your entire workload with Intel TDX + NVIDIA CC.
| Feature | Standard Cloud No Protection | On-Prem GPU Local Hardware | Phala Cloud GPU TEE + CPU TEE |
|---|---|---|---|
| CPU/Memory Protection | None | Physical | Intel TDX |
| GPU Encryption | None | None | NVIDIA CC |
| Full VM Isolation | Standard VM | Physical | TDX + CC |
| Attestation | None | None | Intel + NVIDIA |
| Verifiable Code Integrity | None | Manual | Cryptographic |
| Cloud Convenience | Easy | DevOps burden | Managed |
| Costs | Variable | High CapEx | Pay-as-you-go |
| AI Use Cases | Agent only | Inference/Fine-tune/Train | Agent/Inference/Fine-tune/Train |
What You Can Build with GPU TEE
Real-world applications running on Phala Cloud with complete Intel TDX + NVIDIA Confidential Computing protection
Private Enterprise AI
Train and deploy models on sensitive healthcare, financial, or legal data with complete hardware protection. Your data never leaves the TEE.
User-Owned AI Agents
Build autonomous AI agents that securely manage cryptographic keys and digital assets. Powers platforms like Eliza and Virtuals Game Agents.
ZK Proof Generation
Accelerate zkVM and zkRollup proof generation with GPU TEE. SP1 zkVM runs with <5% TEE overhead—verified with dual attestation.
FHE/MPC Acceleration
Use GPU TEE as 2FA for FHE and MPC systems. Secure key generation, computation integrity, and attestation in one platform. Powers Fairblock and Mind Network.
Multi-Proof Systems
Combine ZK proofs with TEE attestation for double security. Hedge against cryptographic bugs while maintaining verifiability.
Regulatory Compliance
Meet GDPR, HIPAA, and SOC 2 requirements with hardware-backed privacy guarantees. Full audit trail with Intel and NVIDIA attestation.
Three Ways to Deploy GPU TEE
Choose the deployment model that fits your needs—from full control to instant deployment
CVM + GPU: Maximum Flexibility
Deploy your own Docker containers with SSH access to TEE-protected GPUs. Perfect for developers who need complete control.
- ✓Deploy custom Docker containers with full SSH access
- ✓Fine-tune models on private data with complete hardware protection
- ✓Intel TDX + NVIDIA Confidential Computing protection
- ✓Dual attestation reports (Intel + NVIDIA) for verification
What is GPU TEE?
GPU TEE (Trusted Execution Environment) provides hardware-level isolation for your entire AI workload—from CPU to GPU. Unlike traditional cloud computing where your data is exposed to the host system, GPU TEE creates a secure enclave that not even the cloud provider can access.
Phala Cloud is the only platform combining Intel TDX (CPU/memory protection) with NVIDIA Confidential Computing (GPU encryption) for complete Full-Stack TEE protection. This dual-layer security enables private AI training, ZK proof generation, and cryptographic operations with cryptographic verification through dual attestation reports.
REAL-WORLD PERFORMANCE
Hardware Security Without Compromise
Full-Stack TEE protection delivers near-native performance. Benchmarked with SP1 zkVM generating ZK proofs in GPU TEE environment.
Full-Stack TEE Advantage
vs GPU-Only Protection
Performance Benchmarks
TEE Overhead
<5%Near-native performance with complete hardware protection
H200 vs H100
2x FasterUp to 2x faster LLM inference with 141GB HBM3e memory
SP1 zkVM in TEE
4-5% OverheadVerified benchmark generating ZK proofs in GPU TEE
Security Guarantees
Full VM Isolation
Intel TDX Trust Domain isolates entire VM from host
GPU Encryption
NVIDIA CC encrypts all GPU memory during computation
Dual Attestation
Cryptographic proof from Intel + NVIDIA verifying TEE
Compliance Ready
Meets GDPR, HIPAA, SOC 2 with attestation logs
NVIDIA H200
Flagship Performance
The most powerful AI GPU with complete TEE protection:
- 141GB HBM3e Memory
- 4.8 TB/s Bandwidth
- 2x faster vs H100
- Intel TDX + NVIDIA CC
Scale 1-8 GPUs. Deploy instantly or save 27% with reserved pricing.
On-demand from $3.50/GPU/hr or reserved from $2.56/GPU/hr with 6-month commitment.
Learn More About H200 →
NVIDIA H100
Enterprise Standard
Battle-tested GPU for enterprise AI workloads:
- 80GB HBM3 Memory
- 3 TB/s Bandwidth
- 9x faster vs A100
- Intel TDX + NVIDIA CC
Full-stack TEE protection. Perfect for training on private data.
On-demand from $3.08/GPU/hr or reserved from $2.50/GPU/hr with volume discounts.
Learn More About H100 →
NVIDIA B200
Available Now
Next-generation Blackwell architecture with revolutionary performance:
- 192GB HBM3e Memory
- 8 TB/s Bandwidth
- 15x faster inference vs H100
- Intel TDX + NVIDIA CC
Scale 1-8 GPUs. Deploy instantly or save 29% with reserved pricing.
On-demand from $7.99/GPU/hr or reserved from $5.63/GPU/hr with 6-month commitment.
Learn More About B200 →
Common Questions & Answers
Everything you need to know about GPU TEE
What makes Phala Cloud different from other GPU cloud providers?
Phala Cloud is the only provider offering Full-Stack TEE: Intel TDX for CPU/memory protection + NVIDIA Confidential Computing for GPU protection. Most providers offer GPU-only security or no TEE at all, leaving your data exposed to the host system.
What is dual attestation?
Dual attestation means you get cryptographic proof from both Intel (for TDX) and NVIDIA (for Confidential Computing) that your workload runs in genuine TEE hardware. You can verify these attestation reports independently with our open-source tools.
What's the performance overhead of TEE?
Our benchmarks show <5% overhead for most workloads. For example, SP1 zkVM running in GPU TEE shows only 4-5% performance impact compared to non-TEE environments, while providing complete hardware protection.
Can I train AI models on private data?
Yes! Full-Stack TEE protection means your training data, model weights, and inference results are all protected by hardware encryption. Perfect for healthcare, financial, and legal AI applications where data privacy is critical.
How do I deploy my own models?
Three options: (1) Deploy CVM + GPU for full control with SSH access, (2) Use our Confidential AI Models for one-click deployment, or (3) Contact us for Enterprise Solutions with dedicated clusters and custom SLAs.
What about ZK proof generation and FHE workloads?
GPU TEE is optimized for cryptographic operations. We support SP1 zkVM, zkRollups, FHE acceleration, and MPC systems. Our customers include projects like Fairblock, Mind Network, and various zkRollup providers.
Which GPUs are available?
Currently: H200 (141GB HBM3e), H100 (80GB HBM3). Coming soon: B200 (192GB HBM3e). All GPUs include Intel TDX + NVIDIA Confidential Computing protection. Scale from 1 to 8 GPUs per instance.
What's the pricing?
On-demand: H200 from $3.50/GPU/hr, H100 from $3.08/GPU/hr. Reserved (3+ months): H200 from $2.30/GPU/hr (34% savings), H100 from $2.50/GPU/hr (30% savings). Volume discounts available for enterprise.
How do I verify attestation?
Use our attestation API or open-source verification tools. You'll receive attestation reports from both Intel and NVIDIA that you can verify independently. We also provide a public PCCS server at https://pccs.phala.network for quote verification.
