Ex-Nvidia Researcher Sanja Fidler's Veeda AI Bags $90M Seed To Build 'The Matrix' For Physical AI

Toronto-founded Veeda AI has closed $90M+ in seed funding co-led by Khosla and Radical Ventures to build foundation world models for training humanoid robots and autonomous vehicles in simulation.

Ex-Nvidia Researcher Sanja Fidler's Veeda AI Bags $90M Seed To Build 'The Matrix' For Physical AI

Toronto-based physical-AI startup Veeda AI has emerged from stealth with more than $90 million in seed funding to build "the Matrix" for embodied intelligence — a foundation world model designed to let humanoid robots and autonomous vehicles learn through interactive trial and error inside high-fidelity simulated environments rather than the physical world.

One of the largest seed rounds ever raised in Canada

The round was co-led by Radical Ventures and Khosla Ventures, marking one of the largest seed financings in Canadian history and coming just three months after the company was founded. Veeda AI, officially incorporated as Veeda Innovation Inc., is led by chief executive Sanja Fidler alongside chief scientist Huan Ling and chief technology officer Zan Gojcic. All three previously helped build Nvidia's Spatial Intelligence Lab and its first generation of world models for physical AI developers.

Foundation world models as the training substrate for Physical AI

Veeda argues that scaling real-world reinforcement learning for robots is fundamentally impractical: hardware does not scale like GPUs, mistakes carry real physical cost and human-time interaction cannot be parallelised. Instead, the company is developing multimodal foundation world models — generative simulators of the physical world — that can be spun up at data-center scale so that virtual robot bodies can attempt tasks millions of times, fail safely and learn from their own mistakes.

Veeda AI — Scaling Physical AI through Interactive Learning in Simulated Reality

Betting on the "matrix for physical AI" playbook

The thesis puts Veeda in direct company with a growing cohort of world-model contenders backed by the same investors: Fei-Fei Li's World Labs raised $500 million from Radical earlier this year, Waabi Innovation closed a $750 million round to build world models for autonomous trucking and robotaxis (Khosla was a lead), and Decart.ai banked $300 million in May. Veeda's founders say their conviction is that world models will become the "critical infrastructure layer" for every branch of robotics.

Toronto, Zurich, Bay Area, Singapore

Veeda is hiring across four locations — Toronto, Zurich, the San Francisco Bay Area and Singapore — and says it is deliberately staying small and deeply technical. The team plans to combine expertise across simulation, generative AI, 3D graphics, robotics and embodied intelligence to move quickly from research prototypes to a commercially usable simulated-reality platform.

The financing lands during a red-hot summer for physical-AI capital: Prime Intellect closed a $130 million Series A at a $1 billion valuation in July, and robotics venture funding has already crossed $23 billion year-to-date, closing in on the industry's record 2025 total.

Reporting based on coverage from SiliconANGLE, The Logic, BetaKit and The Globe and Mail.

Category: Funding & Investments

Tags: venture capital AI Startups Physical AI world simulation embodied AI robotics funding Seed Funding Nvidia

Related Articles