Closing the Loop on Data Vulnerability
In today's landscape, organizations handle an enormous amount of sensitive data. Protecting this data while it is sitting on a hard drive or moving across a network is standard practice. However, when data is actively being processed in CPU or GPU memory, it is typically decrypted and vulnerable to hypervisor compromises or memory scraping.
Confidential Computing solves this. By leveraging hardware-based Trusted Execution Environments (TEEs), it ensures true end-to-end security—protecting data while it is in use.
Encryption for Stored Data
Encrypting data that's stored on hard drives or solid-state drives (Data at Rest). This prevents physical theft or unauthorized disk access from exposing sensitive information.
Encryption in Transit
Securing data via TLS/SSL as it travels between different systems, clients, and across networks. This protects data from interception and man-in-the-middle attacks.
Encryption in Use
Confidential Computing introduces the final layer: safeguarding data while it is actively being processed in RAM or CPU/GPU registers using hardware-enforced isolation.
Hardware-Enforced TEEs
To implement data-in-use encryption, hardware-based Trusted Execution Environments (TEEs) are created. Enabled securely at the BIOS level, these technologies isolate applications and VMs from the hypervisor, the host OS, and even the system administrator.
AMD SEV-SNP
Secure Encrypted Virtualization with Secure Nested Paging (SEV-SNP) on AMD EPYC™ (Milan/Genoa) processors provides robust isolation for creating Confidential Virtual Machines (CVMs).
- Protects VMs against malicious hypervisors and host OS tampering.
- Hardware-managed encryption keys bound directly to the CPU.
- No code changes required for the applications running inside the CVM.
Intel TDX & SGX
Intel offers two paths: Software Guard Extensions (SGX) for strict application-level enclaves, and the newer Trust Domain Extensions (TDX) available on 4th/5th Gen Scalable and Xeon 6 for full CVM isolation.
- TDX: VM-level isolation, similar to AMD SEV-SNP, requiring zero code rewrite.
- SGX: App-level isolation for executing highly sensitive cryptographic functions.
- Hardware-level remote attestation to verify the environment's integrity.
NVIDIA Confidential Computing (CC)
Historically, GPUs were a blind spot for data-in-use encryption. Today, confidential computing with GPUs pairs host CPUs (Intel or AMD) with secure accelerators like NVIDIA Confidential Computing to encrypt data-in-use across the PCIe bus.
Broadly supported and production-ready, this architecture pairs AMD SEV-SNP or Intel TDX with NVIDIA Hopper (H100/H200) and Ampere GPUs to securely train AI models on highly sensitive data.
The CPU-to-GPU "Secret Handshake"
During boot, the CPU and GPU perform a secure cryptographic key exchange over the PCIe bus, establishing a hardware-encrypted tunnel that cannot be snooped by the hypervisor.
NVIDIA CC-On Mode
Protects data in transit across the PCIe bus and inside the GPU memory using hardware-encrypted bounce buffers and remote attestation without altering native CUDA code.
==============NVSMI LOG==============
Timestamp : Wed Oct 7 15:52:50 2026
Driver Version : 535.104.05
CUDA Version : 12.2
Attached GPUs : 1
GPU 00000000:4B:00.0
Confidential Computing
CC Mode : ON
CC Environment : SEV-SNP
Protected PCIe : Enabled
Bounce Buffers : Encrypted
Deploy Confidential Computing with iDatam
With the utilization of the latest-generation Intel and AMD processors alongside NVIDIA Hopper GPUs, iDatam delivers top-tier performance and zero-trust security on 100% dedicated hardware.
Intel Dedicated Servers
Deploy custom bare-metal servers powered by Intel Xeon Scalable CPUs. Gain absolute BIOS-level control to enable TDX and SGX to build your own fully isolated virtualization environments.
Configure Intel ServerAMD Dedicated Servers
Deploy custom bare-metal servers powered by AMD EPYC CPUs. Leverage SEV-SNP technology at the hardware level to create robust Confidential Virtual Machines (CVM) with hardware-managed encryption.
Configure AMD ServerGPU Dedicated Servers
Train AI models securely using NVIDIA Hopper (H100/H200) architecture. Leverage CC-On Mode with hardware-encrypted PCIe bounce buffers to isolate your proprietary models and datasets.
Configure GPU ServerSuitable Servers for Confidential Computing
Browse our real-time inventory of Intel and AMD servers equipped with hardware-enforced TEE capabilities and optional NVIDIA Confidential GPUs.
No Servers Match Your Filters
Try unchecking "NVIDIA CC GPU Only" or selecting a different country.
Confidential Computing Questions
What is a Confidential Virtual Machine (CVM)? ▶
A Confidential Virtual Machine (CVM) is a highly secure VM encrypted at the hardware level using technologies like AMD SEV-SNP or Intel TDX. The encryption keys are managed by the processor, ensuring that even the cloud provider, hypervisor, or system administrator cannot access or view the data inside the VM while it is running.
What is the difference between Intel SGX and Intel TDX? ▶
Intel SGX (Software Guard Extensions) provides application-level isolation, requiring developers to modify their applications to run inside secure enclaves. Intel TDX (Trust Domain Extensions) operates at the virtual machine level, isolating the entire VM from the hypervisor without requiring any code changes to the applications running inside it.
Do I need to rewrite my code to use NVIDIA CC-On Mode? ▶
No. NVIDIA CC-On Mode protects data in transit across the PCIe bus and inside the GPU memory using hardware-encrypted bounce buffers and remote attestation, all without altering your native CUDA code.
How does the CPU-to-GPU secret handshake work? ▶
Confidential computing with GPUs pairs host CPUs (Intel or AMD) with secure accelerators. During initialization, the CPU and GPU perform a secure cryptographic key exchange (the secret handshake) over the PCIe bus. This ensures all data transferred between the CPU and GPU memory remains encrypted and inaccessible to physical probing or hypervisor snooping.