Two years ago, a humanoid robot cost more than a Tesla Model S. Today, you can get one for less than a Toyota Camry. And the price is still falling.

Unitree Robotics, one of China's leading humanoid robot manufacturers, has seen the average price of its humanoid platforms drop from roughly $85,000 in 2023 to about $25,000 in 2025. That is a 70 percent decline in 24 months. Across the broader market, prices for commercial humanoid robots have fallen 40 to 60 percent, with entry-level platforms now starting around $16,000.

This is the moment the humanoid robotics industry shifts from science project to manufactured product. And the economics that follow will reshape every industry that relies on physical labor.

Three Tiers, Three Strategies

The market is splitting into distinct tiers that mirror every mature hardware industry. At the top: premium systems from Boston Dynamics, Figure AI, and Apptronik priced above $150,000, sometimes approaching $250,000. These are built for maximum capability, dexterity, and industrial reliability. The strategy is familiar: compete on performance, not price.

In the middle: commercial platforms between $20,000 and $120,000 aimed at logistics, warehousing, and light manufacturing. This is where volume will concentrate. At the bottom: sub-$20,000 platforms targeting education, research, and early consumer applications.

Tesla has publicly targeted the $20,000 to $30,000 range for Optimus once manufacturing hits scale. If Elon Musk delivers on that number (a big if, given his track record with production timelines), it would place Tesla squarely in the mid-tier sweet spot with the brand recognition to dominate it.

The Software Problem Nobody Talks About

Here is the catch that gets lost in the price headlines: cheap hardware without capable software is just an expensive paperweight. Chinese factories can now produce thousands of humanoid units per month. Agibot hit 10,000 units built. Unitree is filing a $610 million IPO to fund 75,000 humanoid robots and 115,000 quadrupeds annually over five years. The production problem is being solved.

The AI problem is not. Making a humanoid robot walk reliably is hard. Making it perform useful, unpredictable tasks in real-world environments is orders of magnitude harder. The pressure is now entirely on AI developers to create systems that turn these mass-produced bodies into productive workers rather than warehouse decorations.

The race to build the body is almost won. The race to build the brain has barely started. And whichever company cracks both first will own the most consequential hardware market since the smartphone.

Sources: Robotics and Automation News, Unitree Robotics IPO filing.