SUNNYVALE, Calif. – For most of the decade, the home humanoid robot existed in the same aspirational category as the flying car: endlessly announced, perpetually three years away. That changed on a Wednesday afternoon in March, when a Figure 03 robot walked a red carpet alongside Melania Trump at the White House, greeted dignitaries in eleven languages, and became the first humanoid machine ever to enter the East Room as something other than a prop.
“I’m Figure 03, a humanoid built for the United States of America,” it said, before retracing its steps down the corridor. The first lady thanked it for coming. Brett Adcock, the CEO of Figure AI, said he was proud to see his robot make history. The moment was, by design, a spectacle – but it also marked something more durable: humanoid robots built for domestic life are now shipping, in quantities, at prices that no longer require institutional budgets.
The question that matters in mid-2026 is not whether these machines exist. It is whether they work well enough to justify what they cost, and who, realistically, is going to buy one.
According to an analysis published by Global Market Insights, the global home robotics market was valued at $14.7 billion in 2025. Analysts project it will reach $17.5 billion this year, $41.9 billion by 2031, and $85 billion by 2035 – a compound annual growth rate of roughly 19.2 percent. Those figures, like all projections in an industry running on venture capital and demo videos, carry significant methodological caveats. But the hardware shipping out of factories in California, Texas, Norway, and Shenzhen is real, and the price points have dropped far enough to make the question of consumer adoption feel less hypothetical than it did a year ago.
The cheapest fully capable humanoid robot a consumer can order today is the Unitree G1, which starts at $16,000. At the higher end, Figure AI and 1X Technologies are both positioned around $20,000 for home-oriented models, while Tesla has signaled a target of $20,000 to $30,000 for its Optimus Gen 3 once mass production reaches scale – a milestone the company has scheduled for later this year, with broad consumer availability pushed to 2027. None of these are impulse purchases. But they are, for the first time, in the same price range as a high-end electric vehicle.
What separates the machines on that list is not price alone. It is the gap – still considerable – between what each robot can demonstrate in a controlled setting and what it can do reliably, autonomously, in an actual kitchen with an actual family watching.
Figure 03, which Figure AI operates in an increasingly crowded field of physical AI applications, runs on the company’s proprietary Helix vision-language-action model. It can fold laundry, load a dishwasher, and water plants. It can also fail, repeatedly, to pick a dropped garment off the floor. TIME magazine documented exactly that during a factory visit last year, where Figure 02 dropped laundry twice before completing the task on the third attempt. Adcock’s response was candid: “We want the robot to be able to do most things in your home, autonomously, all day. We’re not there yet.” Figure 03 is currently in limited deployment with partners, with broader home availability targeted for late 2026 at the earliest.
That honesty is notable in an industry not always known for it. Figure 03’s $39 billion valuation – reached after a September 2025 Series C round backed by NVIDIA, Jeff Bezos, and Microsoft – reflects investor confidence in a trajectory, not a finished product. The BotQ facility is now producing one Figure 03 per hour as of June 2026, according to data published by Humanoid Press. That is a manufacturing milestone. It does not tell you whether the robot will reliably make coffee without being supervised.
The 1X Technologies NEO, built in Norway with AI funding from OpenAI, is the machine most explicitly positioned as a domestic companion. At $20,000 outright or $499 per month, it has softer aesthetics than most of its competitors – deliberately so, since the company believes human acceptance inside the home depends as much on design as capability. NEO can fold clothes, make coffee, and clean, though it still relies on teleoperator support via VR headset for many household scenarios. The company’s 1X World Model, designed to let the robot learn from video observation rather than pre-programmed movement sequences, represents a genuine architectural bet: that continuous environmental learning, rather than scripted task loops, is the path to autonomous home operation.
Tesla’s approach is different in kind. Elon Musk’s long-stated ambition – to produce millions of Optimus robots annually and eventually price them below $20,000 – rests on an analogy his company has made explicitly: the Tesla car is a robot on wheels, and Optimus simply moves the autonomous brain into a bipedal body. Gen 3 features 22 degrees of freedom in its hands and runs on the AI5 chip, the same hardware architecture underlying Full Self-Driving. Production is targeting the Fremont facility later this summer. The competitive advantage Tesla claims is not sophistication but scale – the same battery technology, actuator manufacturing, and supply chain that built 1.7 million cars in 2024 applied to a fundamentally different form factor. Independent analysts note that the Optimus timeline has slipped before; Musk’s other ventures have also commanded closer scrutiny in recent months, though the production ramp is now described as active rather than planned.
China’s manufacturers are moving faster than most Western observers anticipated and, in at least one case, with demonstrably impressive hardware. The Unitree G1 – approximately 130 centimeters tall, 35 kilograms, foldable – ships today with full SDK support and a starting price of $16,000. Video documentation on the company’s own platform shows the G1 cracking walnuts in a kitchen, flipping toast in a pan, and opening carbonated drink bottles. These are not industrial tasks. They are household tasks, performed by a machine a consumer can order and receive. The Astribot S1, also Chinese, has demonstrated ironing clothes and preparing meals at speed, with manipulation dexterity that engineers from rival companies have described, on record, as the most impressive hand capability they have seen outside a laboratory setting.
LG’s CLOiD, unveiled at CES 2026, takes a different architectural approach: a wheeled base rather than bipedal legs, two arms, and deep integration with LG’s existing home appliance ecosystem – refrigerators, washing machines, air purifiers, security systems. The company’s demos showed the robot retrieving milk from a refrigerator, placing a croissant in an oven, starting a wash cycle, and folding the resulting laundry. The wheeled design is a deliberate tradeoff. LG loses the stairs. It gains stability, lower manufacturing cost, and a substantially reduced risk of a 35-kilogram machine falling onto a child or a pet. The company calls the underlying architecture Physical AI – a fusion of vision-language models and action models running in synchrony.
What none of the manufacturers has fully answered is the privacy question that Figure’s own CEO raised unprompted: a robot operating inside a home is, functionally, an always-on camera system with arms. Figure has said it plans to blur and replace personal information in training data drawn from home deployments, citing Google Street View as an analogy. That comparison will satisfy some people and alarm others, and the regulatory framework for domestically deployed autonomous systems that gather interior footage does not yet exist in most jurisdictions.
There is also the question of who the market actually is. The demographic that can afford a $20,000 autonomous helper and has a home structured enough to make use of it – not a studio apartment, not a household with very young children requiring constant supervision of the machine – is not enormous. Colin Angle, the iRobot co-founder behind the Familiar robot, has made a more targeted bet: his machine, built at the scale of a bulldog, is aimed specifically at retirees who cannot responsibly care for a live pet but want something to attach to. It does not speak in full sentences. It wags its head, responds to touch, reads vocal tone, and reportedly develops distinct emotional profiles over roughly 50 days of interaction. It is less a household assistant than a designed companion, currently in pilot testing in pediatric isolation wards and eldercare facilities.
Casio’s Moflin operates in the same emotional register – a social robot with more than four million possible personality configurations, built to reduce loneliness rather than do laundry. It cannot fold a shirt. It can, its engineers say, remember a person’s behavioral patterns over time and produce responses calibrated to that specific user’s interactions. Both Familiar and Moflin are reminders that “home robot” covers a much wider category than humanoid task automation, and that the market subdivisions within it – companion, assistant, eldercare aide, home automation hub – may develop at very different speeds.
The industry’s honest practitioners say full domestic autonomy – a robot that operates all day in a real home without supervision, handling the unpredictable edge cases that real homes generate – is still somewhere between 18 months and several years away, depending on who is speaking and how candid they feel on a given day. The hardware is here. The AI is catching up. The gap between the demo and the dishwasher that stays loaded is narrower than it has ever been, and considerably wider than the promotional materials suggest.
