Ex-Nvidia insider turned VC says robotics remains in 'Lego era'

Amulya Vishwanath, who spent seven years leading product marketing for Nvidia’s robotics platforms before moving into venture capital, argues that robotics is still in its “Lego era.” The basic building blocks—actuators, robotic hands, software, and safety standards—remain under reconstruction, she says. That assessment runs counter to the belief that robotics, powered by a general foundation model, is about to have its own ChatGPT moment.

Vishwanath’s view is rooted in her time at Nvidia, where she oversaw marketing for robotics platforms. She saw that today’s robots are assembled from disparate parts, much like Lego bricks. Each component—from motors to grippers to control software—must be integrated, tuned, and validated for specific tasks. There is no universal plug-and-play stack.

In this “Lego era,” actuators vary widely in cost, precision, and durability. Robotic hands are still expensive and fragile. Software frameworks are fragmented across vendors and research labs. Safety standards for human-robot interaction are immature. These gaps make it difficult to deploy robots at scale outside controlled environments.

By contrast, large language models reached a ChatGPT moment because they could be trained on massive text corpora and distributed through simple APIs. Robotics lacks comparable data. Physical interaction data is costly to collect, hardware-dependent, and hard to label. A foundation model alone cannot overcome those physical-world constraints.

The former Nvidia insider turned VC therefore pushes back on the hype that general-purpose robots are imminent. While vision-language-action models and humanoid prototypes show promise, she suggests the underlying stack is not ready. Progress will depend on rebuilding actuators, hands, software, and safety standards.

Vishwanath does not deny advances. She frames current work as necessary infrastructure. Until those building blocks mature, robotics will remain in a Lego-like phase. A true ChatGPT moment may require both better models and a rebuilt physical stack, along with patience and investment in fundamentals.