Robots are waiting for a ChatGPT moment: Nvidia’s Les Karpas explains why at TechCrunch Disrupt 2026

The debut of ChatGPT in November 2022 is widely credited with bringing large language models into everyday attention. Robotics, by contrast, has long been discussed and developed but has yet to produce a single, widely recognized breakthrough that shifts public use in the same way. At TechCrunch Disrupt 2026, Nvidia Inception’s Global Head of Physical AI Les Karpas will lay out why that gap persists and what might finally close it.

Karpas points to a fundamental limitation: there is no internet-scale dataset for physical AI comparable to the text datasets that powered early language models. Even autonomous driving companies such as Waymo have relied on years of accumulated road miles to build their datasets, a scale that robotics broadly has not achieved. In response, a growing set of startups is attempting to recreate that breadth through simulation, synthetic data and foundation models trained across multiple robot platforms.

The Real World AI Stage will explore those approaches with Karpas alongside founders from companies including Shield AI, Colossal Biosciences, FieldAI and Foxglove. Nvidia’s own position — visible in Jensen Huang’s recent keynotes and through its Inception program — gives the company a broad view of the startups and sectors trying to bridge the digital-to-physical divide, from manufacturing and mobility to smart cities and beyond.

Karpas brings a cross-disciplinary background to the conversation, having worked at firms such as Stanley Black & Decker, Intellectual Ventures, Herman Miller, iRobot and Cirque du Soleil and holding roles from manufacturing engineer to startup CEO and corporate venture capitalist. Disrupt attendees can hear his perspective in person at Moscone West from October 13–15; tickets are available now with savings of up to $200 before September 25, and up to 30% off for group bundles.