The year of the humanoid robot: a 2024 recap

I kept a running list all year. Every humanoid robot milestone that seemed meaningful. Not press releases. Not renders. Actual demonstrations of a physical robot doing something new.

Forty-seven entries. In one year.

For context, in 2022 my list had eight entries. In 2023, it had nineteen. The rate isn’t just increasing. It’s accelerating.

Here’s what happened in 2024.

The headlines

Figure AI showed Figure 02 having a conversation with a human while completing physical tasks. The integration of language models with physical manipulation was a first for commercial humanoid robotics. The robot identified objects, reasoned about requests, and acted, all in real-time.

Tesla showed Optimus folding laundry. Slowly, clumsily, but successfully. They also showed autonomous navigation through a warehouse environment. The progress from the embarrassing 2022 stage walk to actual task completion is significant.

Unitree demonstrated the H1 doing martial arts, backflips, and rapid locomotion at speeds that exceeded any previous humanoid robot demo. They also dropped the price to $90,000, making their robot cheaper than many cars.

Sanctuary AI had a robot work an 8-hour shift at a warehouse alongside human workers. Not a demo. A shift. It was slow and needed help sometimes, but it completed the workday.

1X Technologies deployed robots for office cleaning tasks. The robots navigated office environments autonomously, avoided obstacles, and completed assigned tasks without human intervention.

Agility Robotics expanded their Digit deployment in Amazon warehouses. Multiple units, working actual logistics tasks, moving totes from conveyors to shelves.

Boston Dynamics retired hydraulic Atlas and introduced the electric version. Less spectacular, more practical, designed for actual industrial deployment rather than YouTube videos.

The pattern

Each of these milestones shares a common thread: the transition from demonstration to deployment. In 2023, companies showed what their robots could do. In 2024, companies showed where their robots would work.

That’s the shift. Not “look at this cool thing” but “look at this useful thing.” The spectacle era is ending. The utility era is starting.

The funding reflects this. Humanoid robotics companies raised over $5 billion in 2024. Figure alone raised $2.6 billion from Amazon, Microsoft, NVIDIA, and Jeff Bezos. These aren’t demo bets. These are deployment bets.

What surprised me

The speed of convergence surprised me. At the start of the year, the gap between Figure, Tesla, and Boston Dynamics felt large. By December, they’re closer than expected. Figure has better manipulation. Tesla has better manufacturing capability. Boston Dynamics has better locomotion. But each is improving in the areas where they’re weak faster than I predicted.

The Chinese companies surprised me. Unitree, Fourier, and LimX went from relative obscurity to serious competitors in under a year. The price points they’re achieving put pressure on every Western competitor.

The language model integration surprised me. I expected the AI brain part of humanoid robots to be a harder problem. But the same models that chat with you on your laptop turn out to be surprisingly effective at reasoning about physical tasks when connected to robot sensors and actuators. The bridge between language models and physical robots was shorter than I thought.

What didn’t happen

Nobody deployed humanoid robots at scale. “Scale” meaning thousands of units, in production, earning revenue. The deployments are pilots. Dozens of units. Proof-of-concept arrangements with large companies.

Nobody solved the hand problem. Every company I tracked still struggles with fine manipulation. Picking up rigid objects is mostly solved. Handling flexible, fragile, or irregularly shaped objects is still hard.

Nobody achieved the price point for consumer adoption. $90,000 is impressive for a humanoid robot. It’s not something you buy for your house. The consumer timeline is still 10+ years out.

2025 prediction

This is the year of first revenue. At least two humanoid robot companies will generate meaningful commercial revenue from deployed units in real work environments. Not pilot programs. Paid deployments with renewal contracts.

The total number of humanoid robots working in commercial settings will exceed 1,000 by the end of 2025.

I might be wrong. I’m usually too optimistic about robots. But 47 milestones in one year suggests a curve that’s hard to stop.

I’ll keep the list running. Same rules. Only physical demonstrations. Only real robots. No renders, no simulations, no press releases without video evidence.

The count starts at zero for 2025. I don’t expect it to stay there long.

Related thinking: