The battle to build humanoid robots will intensify

Several rival firms believe they are the shape of things to come.

November 12, 2025 03:31 AM | Moss, Norway

Blind gymnast

By Alex Hern, AI writer, The Economist

One

DRESSED IN SENSIBLE Nordic knitwear, Bernt Bornich’s robotic manservant is more stylish than practical. Ask it to fetch a can of Coca-Cola from another room, and more often than not, something will waylay it in the process.

Getting the humanoid robot, called Neo, to work as a useful domestic assistant requires help from a remote human overseer and, ideally, an empty house.

Yet by the end of 2026, Mr. Bornich’s firm, 1x, hopes to have 10,000 of its robots in homes around the world.

Two

That makes 1x one of the more ambitious companies among the many now vying to build humanoid robots. Tesla aims to sell a million of its Optimus robots a year by 2030, but so far the carmaker has only a few of them in operation.

Figure, a Silicon Valley darling, has tested just a handful of robots in a BMW car factory, though it recently announced an impressive-looking new model.

Boston Dynamics has shown off the remarkable parkour abilities of its Atlas robot, but under its latest owner, Hyundai, the firm shows little interest in expanding its operations beyond research.

Three

The market leader is China’s Unitree, which delivered around 1,500 humanoid robots during 2024. But its most advanced model, the H1, which costs $90,000, is still described as a research platform.

By contrast, 1x, based in Norway, is targeting the consumer market with Neo. A sporadically successful robot might sound like a hard sell, given its price tag of $20,000.

Yet despite its limitations, Mr. Bornich’s robot can still be useful. “When you have something that does all your chores, you will never give it back,” he says.

Four

The ultimate goal of those building humanoid robots is not to sell a housekeeper. It is to automate all physical labour. Elon Musk, Tesla’s boss, reckons such robots will outnumber people by 2040. In theory humanoids should fit smoothly into a world built for humans. But just how soon they will be ready for prime time is unclear.

Five

Interact Analysis, an industry analyst, reckons the market for humanoids could eventually be worth $2trn, but says safety concerns and regulation will hamper adoption in the short term.

A robot that makes a mistake can cause physical harm. Lighter, less powerful robots are safer but much less capable. “The technology isn’t there yet,” says Rueben Scriven of Interact. He foresees a “humanoid winter” as funding dries up and ambitions plummet.

Six

Videos of humanoid robots show them performing all kinds of acrobatic feats, but none of them can enter a stranger’s kitchen and make a cup of coffee.

The reason is what Nvidia’s director of robotics, Jim Fan, calls the “blind gymnast” problem. Training robots for thousands of simulated hours in virtual reality can give them incredible athleticism but grants them no understanding of how the world works.

Seven

Humans learn with their whole bodies, not just their brains. Developing “embodied intelligence” will require training a robot brain with a body around it, says Jim Torresen, a roboticist at the University of Oslo, who once taught Mr. Bornich.

And that is why 1x is so eager to get its Neo robots into people’s homes in the coming year—to gather training data about the real world. It hopes this will enable its robots to improve rapidly. Humanoids are on the march. But don’t expect C-3PO in 2026.

Conclusion

To stay on top of the biggest stories in The Economist, please follow the account.

To share with friends

Recommended Reading

• Beijing Delivery Workers’ Autobiography “I Deliver in Beijing” Becomes a Best-Seller Overseas

• Tom Standage Predicts the Top Ten Trends for 2026

• Will the Bottom of the Career Ladder Disappear—Will Humans Be Replaced by AI?