Ex-Nvidia Executive Says Robotics Is Still in Its 'Lego Era'

Amulya Vishwanath, who spent seven years leading product marketing for NVIDIA's robotics platforms, says robotics is currently in its "Lego era" — with the basic building blocks like actuators, robotic hands, software, and safety standards still being rebuilt. Her verdict carries weight: few investors have seen the industry's internals from a comparable vantage point, and she is now putting capital behind that view as founder and managing partner of Techable VC, an early-stage firm investing in physical AI across robotics, materials science, and life sciences.

Her position is a direct counterpoint to one of the most repeated claims in the robotics trade. A vocal camp argues the field, powered by a general foundation model (a single AI system trained broadly enough to handle many tasks, rather than one narrow job), is about to have its "ChatGPT moment" — the industry-shaking inflection that followed the chatbot's launch in late 2022. General Intuition CEO Pim de Witte told TechCrunch's Equity podcast earlier this year that a single general foundation model would make much of today's specialized robotics work redundant, and would constitute exactly that moment.

Vishwanath disagrees, and her framing explains why. The "Lego era" metaphor says the industry is still assembling primitives. Actuators (the components that convert energy into physical movement), robotic hands, software stacks, and safety standards do not yet form a mature, interoperable layer. Until they do, the argument goes, a general model has nothing stable to sit on top of. The claim is not that progress has stalled; it is that the plumbing phase comes before the product phase.

She laid out this view on the latest episode of TechCrunch's Equity podcast, where Dominic-Madori Davis caught up with her to discuss where she is writing her first checks. The stakes are straightforward. If the ChatGPT-moment camp is right, specialized robotics startups risk obsolescence and capital should chase the model layer. If Vishwanath is right, the durable returns sit in the unglamorous components — hands, actuators, safety infrastructure — that everyone else treats as solved.

Who is Amulya Vishwanath?

Vishwanath is one of the first women executives to spin out of Nvidia and launch her own early-stage firm. Seven years inside Nvidia's robotics organization gave her a view of both the hardware pipeline and the developer ecosystem that depends on it. Nvidia sits at the center of that ecosystem, supplying the compute and platform software on which much of modern robotics research runs — which makes an insider-turned-investor's skepticism about imminent generalization a data point worth weighing, not a curiosity.

Her firm, Techable VC, invests in physical AI — a term covering machine intelligence that acts on the physical world — across three areas: robotics, materials science, and life sciences. The breadth itself signals a thesis. Rather than betting on one vertical reaching a breakthrough, she is betting on foundational capability that spans domains.

The caveat writes itself: a fund anchored in components has an interest in the components story. Vishwanath did not, in the released excerpts, name the specific startups receiving her first checks or disclose fund sizes. Readers should treat the "Lego era" framing as an informed investor's thesis, not a settled industry consensus.

What does the disagreement actually turn on?

The split between Vishwanath and de Witte reduces to one question: does capability in AI models transfer cleanly into the physical world? ChatGPT's late-2022 launch demonstrated that a general language model could absorb enough text to converse, code, and reason at a level that stunned users. Language had one enormous advantage — the training data already existed, at internet scale.

Robotics lacks that. Physical interaction data is expensive to collect, hardware varies wildly, and safety failures break objects and people rather than sentences. Vishwanath's "Lego era" observation points at the consequences: without standardized actuators, hands, software, and safety norms, every lab and startup rebuilds the same foundations. That fragmentation, on her account, is why the field's ChatGPT moment has not arrived — and why predicting its imminent arrival understates the engineering left to do.

The bullish camp sees it differently. If a general foundation model for robotics does emerge, it could subsume the specialized work of countless companies, the way ChatGPT compressed years of narrow chatbot development. De Witte's position is that this consolidation is not a risk but the destination — and that the winners of that moment will render redundant much of what today's robotics teams build.

Both camps agree the moment would be transformative. They disagree on the calendar, and on whether the missing piece is a model or a parts bin.

Why it matters for the market

For founders and limited partners, the distinction is money. Capital allocated to specialized robotics applications on the assumption that generalization is imminent looks very different from capital allocated to component and infrastructure layers expected to compound for years. Vishwanath's fund sits explicitly in the second bucket, while de Witte's prediction, if correct, would reward the first.

For the broader AI industry, the robotics debate is a stress test of the foundation-model thesis itself. Language models proved generality in a domain with abundant data and no physical risk. Robotics offers neither. Whether the same architecture paradigm eventually crosses that gap — after the Lego era ends — will shape how far the current AI investment cycle extends beyond software.

Vishwanath's own trajectory adds a market signal of its own: experienced operators leaving platform companies like Nvidia to fund early-stage physical AI suggests confidence that the category rewards dedicated, specialist capital, whatever the timeline. The Equity conversation with Dominic-Madori Davis covers her first-check strategy in full — and until those checks are disclosed, her thesis stands as the sharpest version of the argument that robotics' ChatGPT moment, however real it eventually proves, is still waiting on the blocks.