Former Nvidia AI lab leader raises $90M to build world models for physical AI

From Nvidia's Research Lab to Physical AI

Sanja Fidler, who spent years leading Nvidia's Toronto AI research lab, has left the company to start Veeda AI — a startup focused on building world models, the kind of AI systems that can simulate physical reality well enough for robots to learn from it. The company publicly announced $90 million in seed funding on August 19, 2026, backed by Khosla Ventures and Toronto-based Radical Ventures. The round ranks among the largest seed financings ever raised by a Canadian company, as first reported by BetaKit.

What World Models Actually Are

World models are AI systems that build an internal simulation of how the real world behaves — how objects move, how forces interact, how space relates to time. Rather than learning by doing things in the real world (which is slow and expensive with physical robots), AI agents trained on a good world model can run millions of simulated experiments in software before being deployed to hardware.

Fidler, who also holds a professorship at the University of Toronto, has described world models as «where the next breakthrough lies» in AI. Veeda's stated mission is to «build simulated reality for Physical AI» — not a general world simulator, but a purpose-built infrastructure layer for the robotics industry.

The Team

Fidler is joined by two former Nvidia colleagues: Zan Gojcic as CTO and Huan Ling as Chief Scientist. All three overlapped at Nvidia's Toronto lab, where they worked on 3D understanding, neural rendering, and physics simulation — all directly relevant to the world-model problem.

The company was incorporated in June 2026 as Veeda Innovation, barely two months before the public announcement. The speed from incorporation to $90 million seed reflects both the founders' track record and the current intensity of investor interest in physical AI as a category.

Physical AI Is Heating Up Fast

Veeda's fundraise lands in the middle of a broader surge of investment in physical AI — AI systems designed to interact with the physical world through robots, autonomous vehicles, and industrial automation. The thesis is that large language models solved language; world models will solve physical intuition.

Khosla Ventures and Radical Ventures are both long-time backers of foundational AI research. Radical in particular has a track record of backing Canadian AI researchers at the frontier: the firm has invested in companies spun out of the Vector Institute and the University of Toronto for years. Veeda fits that pattern precisely.

What Comes Next

Veeda has not announced a product, a timeline, or commercial partners. The $90 million will fund model development and hiring. Given Fidler's background in multimodal 3D understanding and the team's depth in physics simulation, the most likely near-term output is a training environment — a simulated world that robotics companies can plug into to accelerate how quickly their systems learn to interact with objects, spaces, and humans.