Tesla Allocates $2B to xAI for 'Physical AI'
Tesla is reportedly allocating $2 billion to its xAI division to accelerate its “physical AI” and embodied intelligence initiatives, even after posting a Q4 earnings beat. The investment aims to master general-purpose robotics at scale, with the Optimus humanoid at the center of the effort. The company is retooling its Fremont facility to support the ramp-up of Optimus for both industrial and consumer applications.
- The $2 billion is part of xAI's larger Series E funding round, which raised a total of $20 billion and valued the AI company at approximately $230 billion. Other prominent investors in this round included Nvidia, Cisco, and Fidelity. - A "Framework Agreement" signed on January 16, 2026, formalizes the collaboration, allowing Tesla to leverage xAI's digital models like Grok to accelerate its own physical AI applications, including Full Self-Driving (FSD) and Optimus. - This investment deepens an existing relationship, as Tesla already supplies Megapack batteries for xAI's data centers, and the Grok chatbot is integrated into some Tesla vehicle infotainment systems. - The move faced some initial shareholder resistance; a 2025 non-binding proposal for Tesla's board to invest in xAI failed to pass, but the company proceeded, arguing it aligns with its long-term strategy outlined in "Master Plan Part IV". - Elon Musk has stated he expects to have thousands of Optimus robots working in Tesla factories by the end of this year and anticipates selling them to the public by the end of next year. - The investment occurs as SpaceX, another Musk-led company, announced its acquisition of xAI on February 2, 2026, creating a combined entity valued at $1.25 trillion and structuring xAI as a wholly owned subsidiary. Tesla's investment now represents an indirect stake in the combined SpaceX-xAI enterprise. - This strategic focus on embodied AI is not unique to Tesla; competitors like Figure AI have raised over $1.9 billion from investors including Jeff Bezos, Microsoft, and NVIDIA, and are building dedicated factories to produce 12,000 humanoid robots annually. - The core technology enabling this push is the advancement of foundation models, which are large-scale neural networks trained on vast datasets that allow robots to generalize learning across various tasks, a significant shift from task-specific programming.
