German electronics company Infineon aims to advance physical AI security against quantum-era threats with a certified module for the Nvidia Jetson Thor platform.

Infineon has integrated its Optiga SLB 9672 trusted platform module (TPM) with Nvidia’s Jetson Thor platform. The hardware-based offering securely stores cryptographic keys and verifies system integrity at the chip level, establishing a certified, quantum-resilient root of trust for physical AI.

The integration strengthens the security foundation, enabling robots and autonomous systems to operate securely and reliably across their full lifecycle. As these systems move from controlled environments into factories and public spaces, the impact of a security failure extends beyond data loss to operational disruption and regulatory liability. For the robotics industry, the security architecture decisions made at design-in have lasting commercial and compliance implications.

“Robots that sense, think and act in the real world are only as trustworthy as the security foundation they are built on,” said Stephan Zizala from Infineon. “Infineon’s Optiga TPM brings a hardware root of trust to the Nvidia Jetson Thor platform that has been proven across hundreds of millions of devices. This integration meets the long-lifecycle and real-time demands of robots operating safely and securely at scale. Post-quantum cryptography designed enables a foundation which remains protected not just for today’s deployments, but for the full life of every robot that relies on it.”

Deepu Talla, vice president of robotics and edge AI at Nvidia (www.nvidia.com), added: “Physical AI systems operate in the real world, where security is foundational. Infineon’s certified Optiga TPM for Nvidia Jetson Thor helps developers protect keys, verify software integrity and securely provision robot fleets at scale, establishing a hardware-based root of trust for secure and resilient autonomous systems.”

The EU Cyber Resilience Act, EU AI Act, IEC 62443 for industrial systems, and sector-specific standards in healthcare and automotive are working towards requirements for demonstrable, auditable security at the hardware level, creating a compliance-driven demand signal that Infineon and Nvidia want to serve.

The Optiga TPM technology provides a physically isolated, FIPS and Common Criteria certified offering that is separate from the application processor. It delivers measured boot and remote attestation, allowing operators and regulators to verify cryptographically at any point in a system’s operational life that its software stack is genuine and unmodified. It also provides hardware-protected storage for proprietary AI model keys, encrypted communications and cryptographically signed over-the-air updates.

Said to be the industry’s first TPM protected by a post-quantum secured firmware update mechanism, it is designed as a root of trust that is protected from being compromised as the cryptographic threat landscape evolves. Developers building physical AI applications on the Jetson Thor platform can rely on the hardware security foundation established at the architecture stage and will remain protected against current and upcoming cryptographic threats on robot systems.

The TPM embeds algorithms including ML-KEM and ML-DSA, standardised by the US National Institute of Standards & Technology (NIST) in 2024. Companies building on the current Optiga TPM today should be able to make an easy transition. For the robotics industry, this matters beyond technical readiness. Regulatory frameworks governing physical AI are already moving in the direction of mandatory PQC compliance, and the architecture decision made at the outset determines whether a deployed robot fleet can meet those requirements across its full deployment period or faces costly hardware intervention when mandates arrive.

Humanoid robots rely on a chain of semiconductor functions to sense, think and act safely and securely, spanning sensing, actuation, power management, connectivity and security. Infineon addresses all these functional blocks through a portfolio of products, with an estimated semiconductor content of approximately $500 per humanoid robot. Security components, including TPMs, represent a growing share of that content as regulatory requirements mature.

A reference design for the SLB 9672 is available. More information is available at www.infineon.com/OPTIGA-TPM-SLB9672, www.infineon.com/pqc and www.infineon.com/cra.

• Infineon and VinRobotics have signed a memorandum of understanding (MoU) to collaborate on the development of humanoid robots. VinRobotics (vinrobotics.net), a Vietnam-based intelligent robotics company, brings regional expertise and industrial scale to the partnership. As part of this collaboration, a joint competency centre will be established at VinRobotics’ Hanoi office, serving as a dedicated hub for research, development and innovation in humanoid robotics.