Why domestic robots are closer than you think

Billions are pouring into humanoid robots, but the most successful domestic robot company just went bust.

Start-ups with billion-dollar valuations are racing to mass-produce domestic robots. Boston Dynamics can do more than backflips now, Tesla’s Optimus keeps popping up, and robot sports are thriving in China. It feels like the robot takeover of our homes is finally imminent.

And yet iRobot, the company behind the Roomba and arguably the most successful domestic robot ever made, just filed for bankruptcy. Its value collapsed from $3.56 billion in 2021 to about $140 million. So which is it: robot revolution or another cycle of hype?

As of 2025, the global home robotics market was worth roughly $15 billion. About 70% of that comes from robot vacuums and mops, now found in 78 million North American homes. These floor cleaners have normalised autonomous machines in our daily lives. Sales are growing at around 17% annually.

Rodney Brooks, the MIT roboticist who co-founded iRobot, thinks the current humanoid hype is overblown. “Today’s humanoid robots will not learn how to be dexterous,” he wrote recently, pointing to the lack of touch data needed to train AI for fine motor tasks. The human hand contains 17,000 mechanoreceptors, and we simply don’t have equivalent training datasets.

Others disagree. Sergey Levine, a robotics researcher at Berkeley, gives a median estimate of 2030 for fully autonomous household robots. His reasoning: unlike self-driving cars, where mistakes are fatal, home robots can afford to fumble. A dropped dish is a learning opportunity, not a catastrophe.

Five years ago, robots were sophisticated puppets following pre-programmed routines. Large language models and vision-language-action models have changed this. Levine’s team at Berkeley has created robots with a “visual cortex” connected to a “motor cortex” through an LLM. When you tell this robot to “clean up the kitchen,” the language model breaks down the task, the vision system identifies objects, and the motor system executes.

These robots now have something resembling common sense. They can look at a “wet floor” sign and understand it means danger without ever having encountered one before. As Levine puts it, “the big benefit that recent innovations in AI give to robotics is the ability to leverage prior knowledge.” And when one robot learns something, they all learn it.

The money flowing in reflects this confidence. Figure AI raised over $1 billion at a $39 billion valuation and is building a factory to produce 12,000 robots per year, with plans for 100,000. Tesla’s Optimus has gone from a person in a robot suit in 2021 to functional prototypes sorting objects and playing board games. Musk claims they’ll cost under $30,000.

Then there’s 1X Technologies, a Norwegian company that has opened pre-orders for a consumer robot called NEO, starting delivery in 2026. The first units come with restrictions and use a hybrid approach where human operators can take control remotely for unfamiliar tasks. When companies build factories before they have customers, it signals a serious bet that the market is coming.

The standard argument for humanoid design is that our homes are built for human bodies. Stairs, door handles, kitchen counters, and light switches all assume a roughly human-shaped operator. A humanoid robot can navigate these spaces without modification.

Brooks predicts the successful domestic robots of the future “will look nothing like humans.” They’ll have wheels instead of legs, multiple arms, and forms optimised for function. The human body is designed for bipedal locomotion and social signalling, not for folding laundry or loading dishwashers.

Dyson, which knows something about domestic robots, is investing in robotic arms rather than humanoid bodies. Amazon’s Astro is essentially a tablet on wheels. Ken Goldberg, a robotics professor at Berkeley, offers perspective: “Humanoid robots are still far behind the hype of AI chatbots, given the complexity of the real world.” Moving objects in unstructured environments is exponentially harder than processing digital information.

Research from the World Economic Forum suggests robots could handle about 39% of household chores within the next decade. The tasks most ready for automation are repetitive and time-consuming: loading dishwashers, folding laundry, tidying up. But the real driver of adoption might be elder care.

Globally, there is a massive caregiver shortage. Countries like Japan don’t have enough young people to care for their ageing population. By 2030, about 25% of elderly individuals living alone could benefit from robot-assisted care. Stanford’s Professor Allison Okamura is developing “soft robots” for caregiving, with compliant arms that can help someone out of bed without bruising them.

Okamura thinks specialised helpers like these will arrive in five to ten years. The general-purpose robot butler might not come until the 2040s. Brooks thinks today’s humanoid efforts will be “long gone and mostly conveniently forgotten” by then.

The probable future involves specialised robots that do specific tasks well rather than general-purpose humanoids that do everything. And those robots might not look anything like us. They might have wheels instead of legs, multiple arms instead of two, forms built for function over familiarity. C3PO might arrive, but he might arrive on wheels.