Enclave

Enclave is a cloud compute platform that runs GPU workloads inside confidential enclaves, meaning the operators physically cannot see your data even though they own the hardware. It pairs flagship NVIDIA GPUs with AMD SEV-SNP trusted execution environments, and every deployment comes with cryptographic attestation you can verify client-side before sending anything sensitive.

The problem it solves is trust in cloud GPU compute. When you rent GPUs from a traditional provider, you're trusting that they won't inspect your data, but that trust is policy-based, not hardware-enforced. Enclave flips that by using confidential computing where the hardware itself prevents the operator from accessing what's running inside. If you're processing proprietary models, private data, or anything you can't afford to leak, this architecture removes the need to just take someone's word for it.

Technically, you deploy WebAssembly applications with WASI HTTP support into sandboxed environments running on confidential GPUs. Each tenant gets process isolation, VRAM is zeroed before reuse between tenants, and TLS terminates inside the enclave so even network traffic is protected end to end. The platform exposes a CORS-enabled API, which means you can hit it directly from browser-based applications without backend proxies.

The audience is developers building privacy-sensitive GPU applications, crypto-native projects that need verifiable security, and enterprises running workloads where compliance requires proving no third party can access the data. If you've ever hesitated to use cloud GPUs because you couldn't fully trust the provider, this is designed for that scenario.

What differentiates it from other GPU cloud providers is the attestation model. Before you send data, you can verify a cryptographic report that proves the enclave is configured correctly and hasn't been tampered with. This isn't a promise in a terms of service but a hardware-backed proof you can check in code. The team also publishes source code on GitHub and maintains open-source verification tools.

Pricing is metered per second and paid in ETH or USDC. GPU and VRAM run about six dollars per hour for a full card, with fractional shares available. CPU and RAM cost around three dollars per hour per node. Payments are forward only, unspent time survives a stop but can't be withdrawn, which is a crypto-native model that matches the wallet-based account system.