Deploy AI ComputeWithout Compromise
On-demand NVIDIA GPUs for training and inference
brew install manifold-inc/tap/targonGPU pricing
Secure compute
On-demand GPUs for Training & InferenceSandboxes for Code and Agents
Virtual Machines
On-demand GPU and CPU VMs with access over SSH. Up to eight GPUs per VM.
Bare Metal
Servers with nothing between you and the hardware.
Sandboxes
Isolated environments for running code and agents.
Reserved Capacity
Reserve GPUs for a fixed term, arranged with our team.
Confidential computing
Confidential VMs on Intel TDXData Protected While in Use
Hardware-isolated trust domains
Confidential TVM VMs run inside Intel TDX trust domains, designed to keep the host OS and hypervisor out of your VM's memory.
Attestation
A VM's disk key is released only after its Intel TDX measurements match expected values. If attestation fails, the disk stays encrypted.
Encrypted in use
Memory is encrypted with keys managed by the CPU, and NVIDIA Confidential Computing encrypts transfers between the VM and supported GPUs.
Boot-time Attestation
Fig.2 Trust boundary
HostOutside the trust boundary
- Host OS
- Hypervisor
- Machine owner
Trust DomainIntel TDX
Memory encrypted with CPU-managed keys
Disk key released only after attestation
GPUNVIDIA
Encrypted transfers
Available on TVM virtual machines running on supported Intel TDX and NVIDIA Confidential Computing hardware. Bare metal servers are not confidential.
Developers
CLI & SDKs for DevelopersAutomate GPUs from Terminal to CI
API keys
Create an API key in the console, then sign in to the CLI or SDK with it.
Service tokens
Organization-owned tokens for CI/CD, so pipelines don't depend on a teammate's personal key.
SSH keys
Register your public keys once and attach them when you create a VM.
Fig.3 Quickstart
Start Building on Targon
Deploy a GPU from the console or the CLI, or talk to our team about reserved capacity.