Confidential Computing Expands from CPUs to GPUs, Containers, and Agentic AI as Enterprise Demand Accelerates

Confidential computing is becoming a foundational security layer for AI, with protection expanding from CPUs into GPUs, multi-GPU environments, containers, and emerging agentic AI workflows, according to global technology intelligence firm ABI Research. The firm finds that GPU-based confidential computing has become a major catalyst for Confidential AI adoption as organizations look to protect data, models, and memory while AI workloads are actively running.

“Confidential computing is no longer a CPU-only conversation,” says Aisling Dawson, senior analyst at ABI Research. “As AI deployments scale and enterprises face growing pressure to secure data in use, the market is shifting toward Confidential AI architectures that can protect inference, training, and increasingly agentic workflows without sacrificing the performance needed for production environments.” ‌

ABI Research notes that CPU-based confidential computing is closest to mass-market adoption, but GPU-based deployments are seeing especially strong momentum because they address one of the biggest gaps in AI security: protecting data while it is being processed. The firm also highlights rising interest in confidential containers alongside confidential virtual machines, as enterprises running cloud-native and Kubernetes-based environments seek lower complexity, reduced attack surfaces, and more flexible deployment models for AI workloads ‌

The vendor landscape reflects that momentum. NVIDIA continues to set the pace in confidential GPUs with advances spanning Hopper, Blackwell, and Vera Rubin platforms, while Intel, AMD, and Arm are strengthening the CPU and heterogeneous compute foundations needed to support broader Confidential AI deployments. On the software side, companies including Anjuna, Red Hat, Fortanix, Decentriq, and IBM are developing platforms and controls that help enterprises secure AI model weights, enable attestation, support confidential containers, and address new agentic AI governance requirements ‌

“Agentic AI is creating a new security inflection point for the confidential computing market,” Dawson says. “While confidential computing will not solve every agentic risk, it is emerging as one of the most credible hardware-rooted approaches for securing agent execution, protecting sensitive memory, strengthening auditability, and building trust in high-value enterprise AI systems. Vendors that pair strong attestation, performance optimization, and ecosystem collaboration will be best positioned to capture this opportunity.” ‌

These findings are from ABI Research’s Confidential Computing and AI report, part of the company’s Quantum Safe Technologies research service, which includes research, data, and ABI Insights.