Fortanix Confidential AI

Fortanix is a data and AI security company headquartered in Santa Clara, California, with offices in Eindhoven, Bengaluru and Singapore. It builds a platform on confidential computing that protects data, AI models and applications while they run, as well as at rest and in transit.

Fortanix Confidential AI runs AI workloads inside hardware trusted execution environments on confidential CPUs and GPUs. It combines Confidential Computing Manager for attestation with Data Security Manager for key release, and in October 2026 added support for CNCF Confidential Containers on Kubernetes. It can be deployed as SaaS, on-premise or fully air-gapped.

• Confidential containers: Kubernetes AI workloads run in TEEs via CNCF CoCo.
• Attestation: Confidential Computing Manager verifies containers are genuine and in a TEE.
• Secure key release: Data Security Manager releases keys only when policy is met.
• Composite attestation: CPU and GPU are verified together.
• High availability: clustered key release avoids a single point of failure.
• Support: SLA-backed 24x7 support.

• Vietnamese cloud and data-center operators offering sovereign AI hosting with verifiable protection.
• Banks and insurers running proprietary models on sensitive customer data in air-gapped environments.
• AI software vendors protecting model IP when deploying inside customer infrastructure.

• Proprietary models and sensitive data stay protected during AI inference.
• Composite attestation verifies the CPU and GPU together before keys are released.
• Deployment options cover SaaS, on-premise and fully air-gapped environments.
• The release cites FIPS 140-2 Level 3, SOC 2 Type 2 and ISO/IEC 27001.

Enterprises building private AI factories and sovereign AI clouds want proof that data is protected during inference, not only at rest or in transit. Confidential computing on CPUs and GPUs addresses this gap. Vietnamese data centers and cloud providers exploring local AI hosting may face similar sovereignty and compliance questions.

Headquartered in Santa Clara, California, with offices in Eindhoven, Bengaluru and Singapore; Confidential AI for NVIDIA-powered AI factories announced in March 2026; partners include NVIDIA, AMD, Intel, Red Hat, HPE and Dell; founding year not stated on the pages fetched.