Tesla Optimus: History & Origins

Milan Kovac, who had led the Optimus programme since 2022, hands over to Ashok Elluswamy, head of the Autopilot team. This signals an even tighter integration between Tesla's vehicle AI and its robotics efforts, reflecting the company's strategy of leveraging shared perception and decision-making systems.

Tesla's custom-built Dojo supercomputer, using proprietary D1 chips, becomes central to training the neural networks driving Optimus's vision and manipulation. Underlying the robot is Tesla's vision-only approach to perception: eight autopilot-grade cameras (no LiDAR) build 3D environmental understanding — the same approach used in Tesla vehicles, simplifying hardware while demanding sophisticated software.

Optimus enters a pilot production phase focused on generating training data for AI systems rather than performing productive commercial work. Units are deployed primarily within Tesla's facilities. Production targets for 2026 discussed in the tens of thousands, with a dedicated facility potentially capable of producing millions annually by 2027. Consumer pricing target of under $20,000 would position Optimus far more affordably than competitors like Boston Dynamics Atlas (estimated $100,000+).

Tesla doubles the dexterity of its hand design. Gen 3 hands feature 22 degrees of freedom per hand, enabling complex manipulation for manufacturing and eventual domestic tasks. Custom rotary and linear actuators span a wide range of force outputs — from 20 Nm for fine wrist movements to 8,000 N for heavy lifting.

Tesla discusses deploying Optimus within its own factories for repetitive tasks like parts sorting, assembly assistance, and quality inspection. External sales to other companies expected to begin once the platform demonstrates reliable autonomous performance. Deployment strategy mirrors Tesla's iterative vehicle development: rapid prototyping, real-world data collection, and software updates to existing hardware.

The Generation 2 design features hands with 11 degrees of freedom per hand — a major leap from initial basic grippers. Video demonstrations show Optimus Gen 2 walking smoothly, performing dance-like movements, and handling delicate objects such as eggs. Weight reduced to approximately 57 kg, improving mobility and energy efficiency.

At Tesla AI Day 2022, the first working, untethered Optimus prototype — nicknamed "Bumble-C" — walks across a stage under its own power and waves to the audience. Though modest compared to established robotics companies' demos, it demonstrated that Tesla was serious about building physical hardware. Milan Kovac begins leading the Optimus programme, overseeing the transition from prototype to pre-production hardware.

Tesla displays a static Optimus prototype at the Gigafactory Texas grand opening, generating public interest and media coverage. The static model previews the intended humanoid form factor before functional hardware became available.

At Tesla's inaugural AI Day, CEO Elon Musk unveils "Tesla Bot" (later named Optimus), with a person in a bodysuit representing the intended humanoid form. Musk outlines a vision for a general-purpose robot controlled by the same AI systems powering Tesla's vehicles, with a projected 20 kg payload capacity. The rationale: Tesla was already building core robotics components — computer vision, neural networks, and high-volume manufacturing — making the leap to robotics a natural extension rather than a fresh start.