Science fiction has long painted vivid pictures of humanoid robots, from the cantankerous Bender in Futurama to the sophisticated Ava in Ex Machina. For a considerable time, it felt as though these mechanical beings belonged exclusively to the realms of screens and literature, a distant aspiration rather than a tangible reality. The prospect of a functional, bipedal robot capable of walking, talking, and performing tasks seemed like a concept confined to fantasy.
However, the landscape is shifting. In the past year, machines have demonstrated a surprising range of abilities, participating in running, boxing, and even playing football at China’s World Humanoid Robot Games. While their performances weren’t always flawless, often involving a misstep or a fall, these events marked a significant step. Concurrently, various companies are actively developing their own lines of humanoids, with a focus on practical utility: assisting with household chores. Towards the close of last year, the robotics firm 1X began accepting pre-orders for its NEO robot, with deliveries scheduled for this year – a development that highlights a crucial milestone in the field. With an increasing number of companies now offering humanoid assistants for purchase, we appear to be entering an era where acquiring a robot butler is a genuine possibility. Prices are no longer astronomical; some options are available for a few thousand pounds, a sum less than that of a modest used car.
This evolving market has prompted the creation of what is likely the world’s inaugural guide for individuals contemplating such a purchase, as well as for those simply curious about the broader implications. This guide aims to go beyond detailing each model’s technical specifications, delving into four essential questions that any prospective robot owner should carefully consider. One certainty remains: potential buyers should proceed with caution. The reality of owning a humanoid robot might encompass far more, and indeed, far less than initially anticipated.
What Humanoid Robots Are Available?
The business of humanoid robots is undeniably generating considerable excitement. Data compiled by Morgan Stanley indicates that over 40 new models were unveiled by companies globally in 2025 alone, with a significant 60 percent originating from China. This surge is partly attributable to the advancements in powerful artificial intelligence systems that are crucial for the robots’ functionality. Investor and tech analyst Cathie Wood has notably described humanoid robots as the most significant avenue for AI integration into our daily lives. Similarly, OpenAI CEO Sam Altman projected in a recent interview that the future will see an abundance of robots performing various tasks on public streets. Cryptocurrency investor Andrew Kang has even predicted that “2026 will be the year of humanoid robots.”
This guide has selected five of the most prominently publicized and accessible models, examining their capabilities in detail. These robots are either currently available for general purchase or are expected to be soon, according to their manufacturers. Importantly, the pricing no longer seems like pure science fiction. Unitree’s entry-level humanoid, the R1, is advertised at just over £4,000. While 1X lists its NEO robot at $20,000 for outright purchase, a more accessible subscription option is available for $499 per month. These figures, while substantial, may appeal to affluent individuals seeking to maximize their productivity by delegating household management. Before making a commitment, however, several critical questions warrant thorough consideration.
Do You Genuinely Need or Want a Humanoid Robot?
At first glance, many of the designs presented by major robot manufacturers appear remarkably similar: a torso, a head, two arms, and two legs. The primary distinctions often lie in their “degrees of freedom,” which refers to the range of motion their articulated joints possess. For instance, 1X’s NEO boasts 22 degrees of freedom in each hand and 75 in total. In contrast, the Figure 03 robot, an iteration of the Figure 02 model already employed in BMW factories for tasks like tidying cupboards and restocking shelves, possesses only 30 degrees of freedom overall.
The emphasis on human-shaped robots is partly driven by practical considerations. Our living and working environments are fundamentally designed around the human form, from the dimensions of stairs and door handles to counter heights and the layout of dishwashers. Nevertheless, this does not always translate to a humanoid being the optimal tool for every specific task. Jonathan Aitken of the University of Sheffield raises a pertinent question: “Do they have to be humanoid?” His response is reflective: “For me, the answer is, ‘No, I don’t think they do.'” While acknowledging the inherent appeal and attention-grabbing nature of a humanoid design, he points out that it is not infrequently the most cost-effective, reliable, or practical choice. He further elaborated, “If I were going to design a cooking robot, it wouldn’t be a humanoid.” Joanna Bryson, associated with the Hertie School in Berlin, shares a similar perspective, finding the desire for a humanoid form “really bizarre.”
For instance, if the primary objective is simply maintaining clean floors, readily available programmable robot vacuums already fulfill this need. Similarly, for tasks involving picking up and storing objects, a bipedal android might be unnecessary. A more suitable alternative could be a mobile platform equipped with a sophisticated robotic arm. The California-based company Sunday offers precisely this kind of solution with a robot named Memo, featuring a humanoid upper body mounted on a wheeled base. The company is currently providing select individuals with the opportunity to test the robot without charge before its broader market release.
Among the most significant cost factors in humanoid robots are the actuators, essential components that convert electrical energy into precise joint movements. These can represent an expense of $500 to $1000 each. Consequently, simpler models with fewer complex limb or appendage designs are likely to be more affordable. Subramanian Ramamoorthy from the University of Edinburgh draws an analogy: “Take a vacuum cleaner and imagine it if it has an arm on it.” He suggests that “That kind of robot is much more likely to enter our homes.”
What Specific Tasks Do You Want Your Robot to Perform?
For those committed to investing in a humanoid robot, Aitken strongly advises a careful evaluation of the intended applications. “The big thing for me is understanding what you want to use the humanoid robot for, because so much other stuff flows down after that,” he explains. Two key factors will significantly determine the extent of its practical assistance: its physical capabilities and its level of intelligence.
Examining the physical aspect first, consider whether assistance is desired for laundry, grocery transport, loading or unloading a dishwasher, or tidying up clutter. While these tasks may sound straightforward, precisely defining their execution can be complex. A helpful approach to assessing what is feasible involves considering the robot’s practical abilities.
Foremost among these is the payload capacity, defining the maximum weight the robot can carry. While stationary, the NEO can lift approximately 70 kilograms, comparable to the weight of some adults. However, its carrying capacity while in motion is limited to about 25 kg, equivalent to a medium suitcase. This surpasses the capabilities of many other models, which might manage to transport individual items like a plate but are unable to handle heavier loads such as substantial bags or piles of laundry. Aitken notes that “The payload is of the order of low single-digit kilograms per arm, which automatically limits the tasks.”
Another significant constraint for current humanoid robots is their energy storage. The battery on the NEO, for example, provides roughly 4 hours of operation on a rapid charge. These batteries are sufficient for less demanding activities like slow movement between rooms and basic tidying. Tasks requiring greater exertion or prolonged operation become more challenging. As Aitken points out, “As soon as I’m starting to ask for it to be more capable and able to do more of the types of tasks that we humans might do, then I’m going to be putting more of a strain on battery life.” Cheaper models inevitably feature less proficient batteries; for instance, the Unitree R1’s battery offers only a 1-hour lifespan.
However, physical strength alone is insufficient for a humanoid robot without corresponding intelligence. These machines are powered by AI models designed to interpret inputs, primarily visual data from integrated cameras, and subsequently orchestrate their movements and actions. Nvidia’s GR00T model, the dominant humanoid robot operating system, is engineered to facilitate seamless interaction with the environment. Nvidia CEO Jensen Huang has previously stated that “the easiest robots to adapt into the world are humanoid robots because we built the world for us.”
Regardless of the specific operating system, the aspiration is for it to be trained on data reflecting human movement patterns, enabling robots to emulate them. Vision-language models, such as those developed by Nvidia and others, are intended to process inputs from the robot’s camera “eyes” and translate them into actionable commands for movement. “World models” also play a role by providing the robot with an internal predictive understanding of its operational environment and its mechanics. These models are utilized to simulate hypothetical scenarios, such as “what happens if I do X?”, allowing for pre-action planning.
While some humanoid robot manufacturers are opting for GR00T as their operating system, others are pursuing independent development. Figure’s robots, for instance, are powered by the company’s proprietary Helix visual-language action model, which functions analogously to a large language model but incorporates visual and verbal inputs. 1X, on the other hand, combines its own internal world model with a vision-language model known as Redwood.
These systems, however, are not without imperfections. Ramamoorthy highlights a widely publicized demonstration of Tesla’s Optimus humanoid in 2024, which was reportedly assisted by human operators using VR headsets to control the robot remotely. Tesla owner Elon Musk has since firmly denied the use of teleoperation in subsequent demonstrations. Consequently, if one anticipates acquiring a plug-and-play android, Ramamoorthy offers a dose of reality: “I think for the average everyday person, I would be really surprised if things started happening very soon.”
This is partly due to the ongoing development of the technology. Early adopters of robotic systems are likely to contribute to their evolutionary process, similar to how data from self-driving vehicle trials informs future improvements. Individuals who have already interacted with some humanoids, such as the NEO, have identified both their strengths and weaknesses. While certain simple tasks could be performed autonomously, many others necessitated remote human guidance, a process known as teleoperation, where a human operator remotely views and directs the robot’s actions through its camera feed. “At the moment, I don’t know what I would do with [a humanoid robot],” states Ramamoorthy, “Because they’re not yet capable enough for doing the things that normal people are interested in.”
Furthermore, privacy concerns are a significant consideration. Unlike the stationary nature of a laptop webcam, the sensors on a robot will be consistently active while it is operational. “On a robot, those things will be active all of the time while it is operational,” notes Aitken, implying that the robot could be passively observing and recording intimate aspects of one’s private life. He strongly encourages prospective buyers to inquire about the data handling practices of any humanoid they purchase. “Who owns that data? Where does it come from? Where does it go?” Bryson goes further, warning, “People think they bought a friend, and they just bought a spy.”
Neither Figure nor 1X provided a response to inquiries regarding criticisms about the data collection methods of humanoid robots. However, companies within this sector typically state that they utilize such data exclusively for performance enhancement purposes.
Will Your Robot Actually Function Reliably, or Last?
Assuming these concerns do not deter a potential buyer, it is essential to critically assess what the advertised upfront cost of a robot truly encompasses. Some robots, such as the Unitree R1, launched in July 2025 priced at £4,400, may appear within financial reach for individuals in higher income brackets. However, purchasing such a unit often provides solely the physical hardware. It does not typically include an integrated AI brain necessary for its operation, making it suitable only for technically proficient individuals capable of installing such software themselves.
In this context, some of the more affordable robots on the market could represent a false economy. It is likely more prudent to view humanoids not as a one-time purchase akin to a dishwasher, but rather as a complex asset like a car or a gaming console. These require ongoing maintenance and software updates to remain in optimal working condition. Therefore, the $499 monthly fee for the NEO might prove a more viable long-term investment than the $20,000 upfront purchase price. Miscalculating this aspect could result in acquiring a robot butler that proves short-lived or entirely inoperable.
What Potential Problems Could Arise?
Finally, beyond contemplating the potential advantages of your prospective robotic assistant, it is crucial to acknowledge the possibility of drawbacks. Aitken states, “Even if you have a task that you do 100 times and 99 per cent of the time you do it successfully, 1 per cent of the time you don’t.” He poses the critical question: “What can happen in that 1 per cent of the time?” He emphasizes that this is “actually the question.”
Put differently, the severity of a robot’s error within a domestic setting warrants careful consideration. While it is impressive that Figure 02 models are contributing to the assembly process at BMW, homes present a far more complex environment for robotic operation compared to factories, which feature predictable workflows and disciplined personnel. A home contains variables like children, pets, clutter, and people moving about absent-mindedly. Aitken mentions that his 18-month-old son moves erratically around their family home. If human parents struggle to consistently track his movements, the prospects for a humanoid robot to do so reliably appear limited.