---
title: "Nvidia’s brain drain is quite deliberate"
description: "Veterans of the chip giant’s robotics labs are founding startups — which will then buy Nvidia chips"
author: "Albion Grey"
published: 2026-09-29T09:00:01Z
modified: 2026-09-29T13:36:44Z
url: https://rews.cc/a/nvidia-s-brain-drain-is-quite-deliberate-7a54b0
language: en
tags: ["nvidia", "robotics", "ai", "startups", "infrastructure", "tech", "business"]
publisher: "Rews (https://rews.cc)"
---

# Nvidia’s brain drain is quite deliberate

*Veterans of the chip giant’s robotics labs are founding startups — which will then buy Nvidia chips*

By Albion Grey · September 29, 2026 · https://rews.cc/a/nvidia-s-brain-drain-is-quite-deliberate-7a54b0

## In brief

- Nvidia veterans including former AI-research vice-president Sanja Fidler are leaving to found robotics startups
- Nvidia has invested in some alumni startups, such as Flexion Robotics and Dyna Robotics, and maintains close ties
- Physical-AI firms raised $33.4bn globally in the first half of 2026, above the $28.7bn raised in all of last year, per PitchBook
- Former Google DeepMind and Waymo staff have also founded robotics startups including Reimagine Robotics and Bedrock Robotics

The best customer a technology company can have is one that used to work for it. Nvidia, which has poured billions into robotics and whose boss, Jensen Huang, popularised the term “physical AI” for machines that interact with the real world, is discovering this happily. The people who built the foundations of that field inside Nvidia are now spinning out to build firms of their own, says Amulya Vishwanath, who left last year to start Techable Ventures, a venture firm focused on physical AI.

The diaspora is impressive. Sanja Fidler, Nvidia’s former vice-president of AI research, left in July to found Veeda AI, which builds simulated worlds in which to train robots — an extension of her work inside the company — alongside other ex-Nvidians. Joel Jang, a former Nvidia researcher, founded Dream Labs, which plies a related trade, the Information reported. Others are tackling everything from software for deploying warehouse robots to machines that carry out physical tasks for businesses.

Nvidia has itself invested in some offspring, including Flexion Robotics and Dyna Robotics, and keeps the apron strings loose rather than cut. Nikita Rudin, Flexion’s chief executive and cofounder, says the chip giant holds biweekly meetings with the startup and grants early access to tools for testing. “We’re still doing the same thing we were doing before,” Mr Rudin said. “We were the beta testers — and now again, just from the outside.”

Why leave so comfortable a perch? Mr Rudin and his cofounder, David Hoeller, departed convinced that robotics had reached an inflection point and was ready for deployment. “We were not necessarily convinced that staying at Nvidia was the fastest way to actually go and deliver that,” Mr Hoeller said. Arsalan Mousavian, another former Nvidia robotics researcher, cites a similar urge; he describes Nvidia as a “hardware-first” company centred on AI chips, and wanted to be “laser focused” on building useful robots. His startup, founded last year, remains in stealth.

Timing helps. Siddhant Haldar, a former Nvidia robotics intern who weighed joining the firm before founding Index Robotics, says the field was long starved of real-world training data. Powerful AI models now pick up some of that slack, and venture funding is booming: physical-AI companies raised $33.4bn globally in the first half of 2026, already surpassing the $28.7bn raised in all of last year, according to PitchBook.

Startups enjoy one more edge over the giants, Mr Haldar argues: they can take risks and deploy imperfect robots, because big companies have “more to lose.” The simulated training worlds in which rivals are investing heavily — AMD recently [paid $8.2bn for Fei-Fei Li’s World Labs](https://rews.cc/a/amd-pays-8-2-billion-for-fei-fei-li-s-world-labs-and-a-place-3a4f9b) — may steady the robots further before they meet the public.

Nvidia is not the only founder factory. Former Google DeepMind engineers have launched Reimagine Robotics and Generalist, and Waymo veterans have built Bedrock Robotics, as researchers trained in fields like autonomous driving pour into robots. But Nvidia’s position is the neatest. The more startups its alumni create, the more chips, software and training kit they will need — and the more Nvidia stands to sell them. The brain drain, viewed from headquarters, looks less like attrition and more like distribution.