Tesla faces employee pushback in Optimus robot training efforts

Tesla logo (public domain) via Wikimedia Commons

Tesla faces employee pushback in Optimus robot training efforts

Workers training their potential robotic replacements have dealt with injuries and grueling physical demands as Tesla scrambles to scale humanoid robot production.

Training a humanoid robot to move like a person turns out to require people to move in ways that aren’t great for their bodies. Tesla employees tasked with generating training data for the company’s Optimus robot have described sprinting, squatting, and performing other physically demanding movements while wearing heavy camera equipment, with some workers experiencing health issues during earlier training phases.

The production push

Tesla has been ramping Optimus production at its Fremont factory using repurposed assembly lines that once built the Model S and Model X. Output has climbed from just dozens of units during a small-scale testing phase in Q2 2026 to several hundred units per week by August 2026.

The target: more than 1,000 units per week by the end of 2026, with a longer-term ambition of roughly 20,000 units per week.

Advertisement

The V3 version of the Optimus hand and forearm contains more than 100 small components that require precise manual assembly and alignment. Most of the components are being sourced from a developing network of suppliers, which means Tesla can’t rely on the kind of mature, battle-tested parts pipeline that supports its vehicle manufacturing.

Training humans to train robots

Tesla has accumulated over 500,000 hours of training data, with plans to double that figure by year’s end. Collecting that data meant employees had to strap on camera rigs and perform physically intense activities.

Some workers experienced injuries and other health issues during earlier training phases. The physical toll prompted Tesla to shift its data collection strategy around June 2025, moving to a vision-only camera recording method that reduced the burden on employees.

Workers are being asked to help build the dataset that could eventually make their own roles obsolete. It’s a dynamic that exists across many industries adopting AI, but it’s unusually literal when you’re physically demonstrating tasks for a robot that’s designed to do those exact tasks.

Leadership turbulence and Musk’s big bet

Milan Kovac, who led the Optimus project, departed Tesla in 2025 amid what has been described as a shift in emphasis toward production design for mass manufacturing.

Musk acknowledged during Tesla’s Q2 2026 earnings call that building an autonomous humanoid robot capable of handling diverse tasks is “one of the hardest things to solve.”

The robots remain primarily tools for internal use and data collection rather than commercial products performing generalized autonomous tasks. Tesla is essentially using its own factories as a testing ground, deploying Optimus units to gather real-world operational data while simultaneously trying to work out the manufacturing kinks.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
Tesla faces employee pushback in Optimus robot training efforts
Tesla faces employee pushback in Optimus robot training efforts

Workers training their potential robotic replacements have dealt with injuries and grueling physical demands as Tesla scrambles to scale humanoid robot production.

Tesla logo (public domain) via Wikimedia Commons

Training a humanoid robot to move like a person turns out to require people to move in ways that aren’t great for their bodies. Tesla employees tasked with generating training data for the company’s Optimus robot have described sprinting, squatting, and performing other physically demanding movements while wearing heavy camera equipment, with some workers experiencing health issues during earlier training phases.

The production push

Tesla has been ramping Optimus production at its Fremont factory using repurposed assembly lines that once built the Model S and Model X. Output has climbed from just dozens of units during a small-scale testing phase in Q2 2026 to several hundred units per week by August 2026.

The target: more than 1,000 units per week by the end of 2026, with a longer-term ambition of roughly 20,000 units per week.

Advertisement

The V3 version of the Optimus hand and forearm contains more than 100 small components that require precise manual assembly and alignment. Most of the components are being sourced from a developing network of suppliers, which means Tesla can’t rely on the kind of mature, battle-tested parts pipeline that supports its vehicle manufacturing.

Training humans to train robots

Tesla has accumulated over 500,000 hours of training data, with plans to double that figure by year’s end. Collecting that data meant employees had to strap on camera rigs and perform physically intense activities.

Some workers experienced injuries and other health issues during earlier training phases. The physical toll prompted Tesla to shift its data collection strategy around June 2025, moving to a vision-only camera recording method that reduced the burden on employees.

Workers are being asked to help build the dataset that could eventually make their own roles obsolete. It’s a dynamic that exists across many industries adopting AI, but it’s unusually literal when you’re physically demonstrating tasks for a robot that’s designed to do those exact tasks.

Leadership turbulence and Musk’s big bet

Milan Kovac, who led the Optimus project, departed Tesla in 2025 amid what has been described as a shift in emphasis toward production design for mass manufacturing.

Musk acknowledged during Tesla’s Q2 2026 earnings call that building an autonomous humanoid robot capable of handling diverse tasks is “one of the hardest things to solve.”

The robots remain primarily tools for internal use and data collection rather than commercial products performing generalized autonomous tasks. Tesla is essentially using its own factories as a testing ground, deploying Optimus units to gather real-world operational data while simultaneously trying to work out the manufacturing kinks.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.