In this episode of The Innovators, I spoke with Pablo Zegers, Chief AI Officer at Kaspix, a company that’s turning everything we think we know about artificial intelligence upside down. Zegers and his team are building analog AI—systems that don’t need transistors, GPUs, or data centers. Their technology runs intelligence directly in hardware, even when that hardware has physical imperfections.

Pablo describes it this way: they’ve found a way to make AI work inside the circuitry itself. Instead of relying on digital computation, their system performs the same matrix calculations that underpin modern AI through electrical relations—voltage, current, resistance—executed at the speed of light. It’s artificial intelligence without the silicon.

The implications are enormous. Imagine a world where every sensor in a car, every industrial machine, or even every household device could process information locally, without having to beam endless data to the cloud. In Zegers’ words, “you could have a tire that knows when it’s about to fail.” By putting intelligence on the edge—literally in the sensor—systems can predict, react, and self-correct without lag or dependence on distant servers.

This isn’t theory. Kaspix is already working with design partners to integrate the technology into real-world applications. The promise goes beyond efficiency and energy savings. If AI runs locally in analog hardware, it becomes nearly impossible to hack, since the intelligence lives in static circuits rather than a centralized, updatable database. In that world, your personal AI could live on your desk—private, secure, and entirely yours.

Zegers has been chasing this vision for four decades, and it’s finally real. As he puts it, “In fifty years, all artificial intelligence will be run on this kind of technology.”