Prem Enclave
The encrypted compute layer for sovereign AI. Run frontier models on GPUs that stay encrypted end to end and verifiable at every layer, on-premise, multi-cloud, or VPC.
Bring Your Own Compute (BYOC)
Your hardware. Our hypervisor.
Enclave is a confidential hypervisor that installs directly on your existing GPU clusters, no hardware swap, no vendor lock- in. One license turns your infrastructure into a verifiable confidential computing environment, encrypted end to end at the GPU. Even Prem can't access your data.
Multi-GPU Attestation: Every model is cryptographically validated before execution, with hardware-verified trust across entire distributed GPU clusters. You know exactly what's running, on every GPU.
Post-Quantum Secure: XWing hybrid key exchange (X25519 + ML- KEM 768) protects payloads against both classical and quantum-era threats. Quantum-safe today, built for government adoption.
Memory Isolation: Payloads are decrypted only inside the enclave. The host OS, hypervisor, datacenter operators, and Prem itself never see plaintext. Nothing is stored. Nothing leaves the enclave.
Encrypted Storage: Ephemeral and persistent block storage secured through an attestation-gated Key Broker Service. Keys are bound to verified enclaves and never exposed to the host.
Runs where you do
Enclave runs in every environment: bring it fully on-prem, host it in your VPC, or consume it directly through the Enclave API. The security model is identical everywhere.
On-premises
Install directly onto your existing racks. Your data stays in your building, and your models run at bare-metal speed with zero external dependencies.
Public cloud (GCP, AWS, Azure)
Same hypervisor, hosted on your cloud account. A single license activates enclave-grade isolation on major providers. You keep your cloud relationship; Prem secures the execution layer.
Private cloud
For regulated environments with custom hosting requirements. Prem integrates with your private cloud provider and maps to your specific jurisdictional and compliance boundaries.
Why license Enclave
Compliance from day one
Provable data isolation, attestation, and audit infrastructure built into the stack. Ready for healthcare, finance, and government mandates.
Own your economics
Fixed costs, full utilization, and a TCO that drops every month you operate. Open- weight models deliver frontier performance at roughly a tenth of closed-API cost, and you stop renting compute you could own.
Weeks to production
Installs on standard GPU hardware. No custom silicon, no long procurement cycles. Operational in 2 to 4 weeks.
Predictable economics
Cloud GPU costs scale with usage; margins shrink as you grow. On-premise flips that: fixed costs and 40-60% TCO reduction within 18 months.
Jurisdictional control
Your data stays in your building. No cross-border transfer risks, no access policies that change with someone else's terms-of-service update.
Bare-metal performance
No shared hardware, no noisy neighbors, no throttling. Every FLOP your GPU delivers, your models actually use.
Architecture
Inside Enclave
Enclave orchestrates CPU and GPU clusters with silicon-level isolation and cryptographic attestation on every execution cycle. Where hyperscaler confidential computing stops at a single GPU, Enclave secures entire clusters, so 100B+ open models run fully private. Built on Intel TDX, AMD SEV, and NVIDIA Confidential Computing.
