Ashish Kumar led Tesla's Optimus AI team for two years. In September 2025, he announced he's joining Meta as a research scientist.
This isn't just another tech job change. Kumar's team replaced Tesla's classical robotics stack with reinforcement learning and video-based training. Those methods gave Optimus the ability to handle tasks autonomously, demonstrated in Tesla's 2025 showcases.
Meta has been aggressively hiring robotics talent from Tesla and other competitors. The company is reportedly offering compensation packages worth "hundreds of millions" for top researchers. Meta established a new robotics division within Reality Labs specifically to develop AI-powered humanoid robots using its Llama foundation model.
The strategic approaches differ completely. Tesla uses third-person video data from 8+ million vehicles to train Optimus. The company aims to produce thousands of robots by 2025, eventually ramping to 50,000-100,000 by 2026. Elon Musk claims 80% of Tesla's future value will come from Optimus and related AI.
Meta focuses on software and first-person data. The company is developing wearable devices (Project Aria, Ray-Ban Meta glasses) to collect AI and robotics training data. Morgan Stanley analyst Adam Jonas predicts Meta could have 20 million glasses in use within two years, creating a massive dataset for training humanoid avatars.
The talent migration signals where momentum is shifting. Milan Kovac, head of Tesla's Optimus program since 2022, resigned in June 2025. Tesla's Optimus AI lead Ashish Kumar left for Meta in September. Meanwhile, Tesla posted 68 new job openings for Optimus in February 2025, seeking mechanical engineers, robotics specialists, and thermal engineers across California, Texas, Utah, and Nevada.
The numbers tell the competitive story. Tesla plans 10,000 Optimus robots by end of 2025. Musk admitted during earnings that his projections sound "absolutely insane" but stands by the timeline. Figure AI raised over $1 billion at a $39 billion valuation with backing from Microsoft, OpenAI, NVIDIA, Amazon. Meta hasn't disclosed specific investment figures but is working on proprietary humanoid hardware targeting household applications.
Goldman Sachs estimates the global humanoid robot market could reach $38 billion by 2035. That's creating a war for the few hundred engineers worldwide who understand how to build these systems at scale.
Design lesson: The winner won't be determined by who builds the best hardware. It will be determined by who has the best training data pipeline and can attract the talent to operationalize it. Tesla has vehicle data and manufacturing experience. Meta has AI research depth and billions of users generating behavioral data. Both are betting they can translate their existing advantages into humanoid robotics leadership.
The companies losing engineers should worry. Talent follows momentum. When your AI lead leaves for a competitor, it signals where researchers think the breakthrough will happen.
Next: The $5.8B robotics market nobody's talking about
Forward this to someone building robotics products or managing autonomy teams.