Tech news in 3 minutes

Embeddedblog_202603

35 d ago

The article discusses the growing trend of processing data at edge nodes instead of in the cloud to reduce latency, improve privacy, reliability, and lower costs. This shift is driven by physical AI domains, particularly automotive and robotics, which require real-time decision-making with low compute costs. Silicon providers are redesigning chips to address these needs, creating platforms embedded in intelligent machines—referred to as "embodied agents"—that act within physical or simulated environments. Classic examples include automotive ADAS, digital cockpits, Industry 4.0, and consumer robots, with humanoid robots emerging as a new growth engine. These systems demand unprecedented compute, sensing, and integration, leading to increased need for heterogeneous compute, custom accelerators, advanced packaging, and high-bandwidth interconnects. SoC developers have access to various silicon solutions, including DSPs, NPUs, chiplet platforms, interconnect and security IP, and design services. These are key to automotive autonomy, digital cockpits, and safety, as well as enabling robotics growth. Real-time perception and sensor fusion rely on multi-modal sensing (vision, radar, lidar, thermal) combined with neural network workloads, pushing inference from cloud to edge. Voice is emerging as a new UI, adding to real-time processing for better user experience and safety in automotive and robotics. Intelligent processing elements like DSPs and NPUs are crucial in heterogeneous SoCs for optimized operations. In digital cockpits, voice UI allows human-machine interaction without screens or apps, using less data than vision systems. It can control infotainment, load profiles, personalize settings, and enable health monitoring or safety features like detecting blind spots or emergency sirens. Voice UI is also useful in consumer robotics. The automotive and robotics segments are poised for growth as they solve real-world problems, with edge AI enabling new use cases. The article concludes that various solutions are available to SoC developers to simplify physical AI deployments. The article originally appeared in the Outlook 2026 and EE Awards 2025 Highlights Issue of EE Times, authored by Amit Kumar, Director of Automotive Product Management for Tensilica DSPs at Cadence Design Systems Inc.