Confidential Cloud

Run AI workloads on TEE-backed infrastructure with end-to-end attestations of hardware-enforced integrity, privacy and verifiability. No code changes. No TEE expertise required.

If you'd rather deploy on your own infrastructure, see Components. For detailed rates, see Pricing. For licensed on-prem deployments, see Enterprise.

Table of Contents

Confidential Inference

Private inference as an API. Pay per token.

Send requests to open-weight models running inside TEEs on our cloud. Your prompts, responses, and model interactions are never visible to us or our infrastructure. Every response includes an attestation proof.

OpenAI-compatible API. Drop-in replacement for existing inference providers. Switch your base URL and get hardware-enforced privacy with no other code changes.

ModelBest for
GLM 5.1Reasoning, multilingual
Qwen 3.5 35BGeneral purpose
DeepSeek V3.2General purpose, coding, long context

Less than 5% latency overhead versus non-confidential inference. See inference pricing for per-token rates. Model requests: founders@confidential.ai.

Confidential VMs

Dedicated VMs on our cloud, running inside TEEs. You rent the VM, we run the infrastructure.

GPU VMs. Single-GPU and multi-GPU configurations for inference, training, fine-tuning, and containers.

GPUVRAMHost CPU TEEBest for
RTX PRO 600096 GB GDDR7AMD SEV-SNPLarge-model inference, high VRAM capacity
H10080 GB HBM3AMD SEV-SNP or Intel TDXTraining, fine-tuning, latency-sensitive inference
B200192 GB HBM3eIntel TDXFrontier training, maximum performance

Configurations. Three confidential computing deployment modes, depending on GPU and workload:

ModeWhat it isRTX PRO 6000H100B200
Single GPU pass-throughOne GPU attached to one Confidential VM
Protected PCIe modeMultiple GPUs share one confidential domain over PCIexx
Multi-GPU pass-throughMultiple GPUs attached to one Confidential VM, each independently attestedxx

In Protected PCIe mode, GPU-GPU communications over the NVLink or NVSwitch interconnect are not encrypted. For Multi-GPU pass-through, GPUs that are part of the same CVM can communicate peer-to-peer over encrypted NVLink connections.

CPU VMs. TEE-backed vCPUs for general-purpose confidential workloads. AMD SEV-SNP and Intel TDX available.

See Confidential VM pricing for per-GPU-hour and per-core-hour rates.

Attestable Builds

Cryptographic proof of what was built and from which source. Kettle, our attestable build service, runs your build process inside a TEE and emits a signed attestable build linking the git commit to the final artifact — no deterministic compilers required.

Every build produces a verifiable, tamper-evident chain of custody: signed attestation of the build environment, provenance for every input, and a hardware-rooted measurement of the output. Downstream consumers can verify the artifact came from the claimed source, built with the claimed toolchain, without needing to trust anyone.

Hardware enforcement of the build pipeline: MAC policies, seccomp filters, and process isolation keep the build environment unchanged between source checkout and artifact emission. Achieves SLSA Build L3.

Connects to your GitHub repo. On every commit: checkout, build inside TEE, signed provenance.

See build pricing for per-minute rates.