Confidential Pods
A pod on VoltageGPU is an Intel TDX enclave with dedicated GPU access. Memory is encrypted, PCIe is protected and deployment is attested, data is isolated from the host operator and the hypervisor by design.
What is a Confidential Pod?
A confidential pod is a computing environment with dedicated GPU access running inside an Intel TDX enclave, memory is encrypted and attestation is verifiable.
Pod Architecture
Each pod runs on Intel TDX hardware with direct access to NVIDIA GPUs. Memory inside the enclave is AES-encrypted, PCIe traffic to the GPU is protected, and every deployment produces a fresh hardware attestation.
Dedicated GPU Access
Full or fractional GPU allocation with CUDA, cuDNN, and TensorRT pre-installed.
Hardware Isolation
Intel TDX encrypts pod memory in use; the host operator and hypervisor cannot inspect it.
Root Access
Full root access via SSH or web terminal. Install any packages you need.
Persistent Storage
NVMe storage persists across restarts. Attach external volumes for large datasets.
Available Confidential GPU Types
Every GPU runs inside an Intel TDX enclave with hardware attestation. Pick the class that fits your confidential workload.
| GPU Model | VRAM | Best For | Starting Price |
|---|---|---|---|
| H200 | 141GB HBM3e | Flagship, large confidential models, multi-GPU | Live on voltagegpu.com/pricing |
| H100 80GB | 80GB HBM3 | Confidential transformer inference, FP8 | Live on voltagegpu.com/pricing |
| RTX PRO 6000 Blackwell | 96GB | Single-GPU confidential workloads | Live on voltagegpu.com/pricing |
Live inventory & pricing
Every pod runs inside an Intel TDX hardware enclave. Query live inventory and authoritative per-hour prices with GET /api/volt/machines, what you see on the dashboard is the same feed.
Quick Start
Deploy your first GPU pod with these steps.
1. Create an Account
Sign up at voltagegpu.com/register and deploy from the console or the API.
2. Add SSH Key
Navigate to Dashboard → SSH Keys and add your public SSH key for secure access.
3. Browse Confidential Inventory
Go to Browse Pods or call GET /api/volt/machines to see live Intel TDX inventory, per-hour prices and available resource_name values.
4. Deploy
Pick a resource, choose a Docker image (or reuse a saved template), attach your SSH key, and deploy. Your balance is debited one hour upfront.
5. Connect via SSH
ssh <workload-uid>@ssh.voltagegpu.comWeb Terminal Available
Access your pod directly from the browser at Your Pods → [Pod Name] → Terminal without needing a local SSH client.
Frequently Asked Questions
What happens if my pod crashes?
Your local volume data (/root) remains intact. External volumes (/mnt) are always preserved. You can restart the pod or deploy a new one and restore from backup.
Can I use custom Docker images?
On standard pods, yes: pass a public linux/amd64 image in the image field of POST /api/volt/pods, with an optional start_command, ports and env. See the API reference. Confidential VMs boot the provider's Ubuntu 24.04 image, with root access to install what you need.
How is billing calculated?
Billing is per-hour based on the GPU type and number of GPUs rented. Billing starts when the pod is deployed and stops when terminated. Stopped pods do not incur GPU charges.
Can I attach multiple GPUs?
Yes, if the node supports it. When renting, you can select the number of GPUs (up to the node's total). CPU, memory, and storage scale proportionally with GPU count.
Is my data secure?
Each pod runs in an isolated container with its own network namespace. Storage is encrypted at rest. SSH access requires your private key. We never access your data.
What software is pre-installed?
Templates include CUDA, cuDNN, Python, and framework-specific packages (PyTorch, TensorFlow, JAX). You have root access to install anything else via apt or pip.
Ready to Deploy Your First Pod?
Create an account to deploy your first Confidential VM.