By Byron V. Acohido

One of the more surprising — and least visible — frontiers of artificial intelligence today is unfolding at the extreme edges of our hyper-connected systems.

Related: AI adoption outpacing controls

Think sensors in forests that detect illegal logging. Smart speakers that recognize the sound of someone falling. Microcontrollers inside doorbells that run face recognition without ever sending data to the cloud.

In each case, AI isn’t happening in a data center or even at the application layer. It’s happening on the chip — embedded directly into the device. And that shift is forcing a rethink of how we design, power, and secure the next wave of connected systems.

At OktoberTech 2025, I sat down with Thomas Rosteck, division president of Connected Secure Systems at Infineon Technologies, to unpack the implications. Infineon sits at the silicon layer, supplying the secured microcontrollers now capable of running machine learning workloads on-device.

Rosteck broke down why this shift matters. Processing data at the edge reduces latency — crucial for real-time decisions. It cuts power consumption — often by more than 90% compared to cloud inference. And it introduces new risks — including model theft, tampering, and privacy exposure — that must now be addressed directly in hardware.

Much of the hard work enabling this transition is happening out of the spotlight — deep inside the semiconductor industry. As showcased at OktoberTech, vendors like Infineon are quietly laying the groundwork for secured, resilient AI systems from the silicon up.

Give a listen to the full Fireside Chat podcast, part of Last Watchdog’s ongoing series on cybersecurity leadership.

Acohido

Pulitzer Prize-winning business journalist Byron V. Acohido is dedicated to fostering public awareness about how to make the Internet as private and secure as it ought to be.

(LW provides consulting services to the vendors we cover.)

November 19th, 2025 | Fireside Chat | Top Stories