LIVE: NXP CEO Rafael Sotomayor speaks at COMPUTEX

Three-layer distributed intelligence system

The architecture separates reasoning (planning), coordination (balance/dynamics), and reflex (motor control) into independent but coordinated layers to eliminate single points of failure and ensure deterministic safety.

Sub-40 millisecond reflex requirements

Physical AI requires extreme edge latency, such as 20-millisecond glass-to-glass response for drones and sub-40-millisecond recovery times for humanoids recovering from bumps, processed locally without cloud calls.

Scalable silicon implementation

NXP deploys this via specific hardware including the 5nm S32N processor family for central vehicle compute and S32K5 products for mission-critical automotive functions like braking and suspension.

World models inject physics knowledge

Instead of dangerous trial-and-error learning, robots use 'world models' to understand concepts like gravity, inertia, and friction—essentially downloading experience without physical mistakes.

VLA models bridge perception and action

Vision-Language-Action (VLA) models create the critical link between what robots see, what they comprehend, and how they move, enabling safe handling of fragile objects like liquid-filled bottles.

Edge model optimization toolkit

NXP's EIQ software compresses cloud-trained AI models through quantization and pruning to run on constrained edge hardware with strict power and memory budgets.

Zero-time trust engineering

Unlike human relationships, machine trust cannot be earned over time; it must be built-in from moment zero using hardware-level security, redundancy, and post-quantum cryptography.

Contain-protect-verify-adapt framework

NXP isolates failures through containment, protects execution via hardware-injected security, verifies safety through ASIL certifications, and adapts to evolving threats via secure over-the-air updates.

No undo button reality

Physical AI failures cause irreversible harm like broken bones or collisions, requiring systems designed to 'go right' even when components fail under Murphy's Law conditions.

Proven productivity metrics

Factory deployments of physical AI robots demonstrate 40% productivity gains over traditional automation, while healthcare diagnostics and lab robot sales surged 610% in 2025.

Strategic industry partnerships

NXP collaborates with Boston Dynamics on cooperative factory robotics and GE Healthcare on precision anesthesia systems, embedding intelligence at the edge of critical medical devices.

Taiwan ecosystem collaboration

Sotomayor emphasized that NXP's ability to commercialize physical AI relies entirely on co-creation with Taiwanese manufacturing partners and global customers who scale the technology.