It takes more than building a humanoid robot to build a humanoid robot product.
Over the next several years, humanoid robots will change the nature of work. Or at least, that’s what humanoid robotics companies have been consistently promising, enabling them to raise hundreds of millions of dollars at valuations that run into the billions.
Delivering on these promises will require a lot of robots. Agility Robotics expects to ship “hundreds” of its Digit robots in 2025 and has a factory in Oregon capable of building over 10,000 robots per year. Tesla is planning to produce 5,000 of its Optimus robots in 2025, and at least 50,000 in 2026. Figure believes “there is a path to 100,000 robots” by 2029. And these are just three of the largest companies in an increasingly crowded space.
Amplifying this message are many financial analysts: Bank of America Global Research, for example, predicts that global humanoid robot shipments will reach 18,000 units in 2025. And Morgan Stanley Research estimates that by 2050 there could be over 1 billion humanoid robots, part of a US $5 trillion market.
But as of now, the market for humanoid robots is almost entirely hypothetical. Even the most successful companies in this space have deployed only a small handful of robots in carefully controlled pilot projects. And future projections seem to be based on an extraordinarily broad interpretation of jobs that a capable, efficient, and safe humanoid robot—which does not currently exist—might conceivably be able to do. Can the current reality connect with the promised scale?
What Will It Take to Scale Humanoid Robots?
Physically building tens of thousands, or even hundreds of thousands, of humanoid robots, is certainly possible in the near term. In 2023, on the order of 500,000 industrial robots were installed worldwide. Under the basic assumption that a humanoid robot is approximately equivalent to four industrial arms in terms of components, existing supply chains should be able to support even the most optimistic near-term projections for humanoid manufacturing.
But simply building the robots is arguably the easiest part of scaling humanoids, says Melonee Wise, who served as chief product officer at Agility Robotics until this month. “The bigger problem is demand—I don’t think anyone has found an application for humanoids that would require several thousand robots per facility.” Large deployments, Wise explains, are the most realistic way for a robotics company to scale its business, since onboarding any new client can take weeks or months. An alternative approach to deploying several thousand robots to do a single job is to deploy several hundred robots that can each do 10 jobs, which seems to be what most of the humanoid industry is betting on in the medium to long term.
While there’s a belief across much of the humanoid robotics industry that rapid progress in AI must somehow translate into rapid progress toward multipurpose robots, it’s not clear how, when, or if that will happen. “I think what a lot of people are hoping for is they’re going to AI their way out of this,” says Wise. “But the reality of the situation is that currently AI is not robust enough to meet the requirements of the market.”
Bringing Humanoid Robots to Market
Market requirements for humanoid robots include a slew of extremely dull, extremely critical things like battery life, reliability, and safety. Of these, battery life is the most straightforward—for a robot to usefully do a job, it can’t spend most of its time charging. The next version of Agility’s Digit robot, which can handle payloads of up to 16 kilograms, includes a bulky “backpack” containing a battery with a charging ratio of 10 to 1: The robot can run for 90 minutes, and fully recharge in 9 minutes. Slimmer humanoid robots from other companies must necessarily be making compromises to maintain their svelte form factors.
In operation, Digit will probably spend a few minutes charging after running for 30 minutes. That’s because 60 minutes of Digit’s runtime is essentially a reserve in case something happens in its workspace that requires it to temporarily pause, a not-infrequent occurrence in the logistics and manufacturing environments that Agility is targeting. Without a 60-minute reserve, the robot would be much more likely to run out of power mid-task and need to be manually recharged. Consider what that might look like with even a modest deployment of several hundred robots weighing over a hundred kilograms each. “No one wants to deal with that,” comments Wise.
Potential customers for humanoid robots are very concerned with downtime. Over the course of a month, a factory operating at 99 percent reliability will see approximately 5 hours of downtime. Wise says that any downtime that stops something like a production line can cost tens of thousands of dollars per minute, which is why many industrial customers expect a couple more 9s of reliability: 99.99 percent. Wise says that Agility has demonstrated this level of reliability in some specific applications, but not in the context of multipurpose or general-purpose functionality.
Humanoid Robot Safety
A humanoid robot in an industrial environment must meet general safety requirements for industrial machines. In the past, robotic systems like autonomous vehicles and drones have benefited from immature regulatory environments to scale quickly. But Wise says that approach can’t work for humanoids, because the industry is already heavily regulated—the robot is simply considered another piece of machinery.
There are also more specific safety standards currently under development for humanoid robots, explains Matt Powers, associate director of autonomy R&D at Boston Dynamics. He notes that his company is helping develop an International Organization for Standardization (ISO) safety standard for dynamically balancing legged robots. “We’re very happy that the top players in the field, like Agility and Figure, are joining us in developing a way to explain why we believe that the systems that we’re deploying are safe,” Powers says.
These standards are necessary because the traditional safety approach of cutting power may not be a good option for a dynamically balancing system. Doing so will cause a humanoid robot to fall over, potentially making the situation even worse. There is no simple solution to this problem, and the initial approach that Boston Dynamics expects to take with its Atlas robot is to keep the robot out of situations where simply powering it off might not be the best option. “We’re going to start with relatively low-risk deployments, and then expand as we build confidence in our safety systems,” Powers says. “I think a methodical approach is really going to be the winner here.”
In practice, low risk means keeping humanoid robots away from people. But humanoids that are restricted by what jobs they can safely do and where they can safely move are going to have more trouble finding tasks that provide value.
Are Humanoids the Answer?
The issues of demand, battery life, reliability, and safety all need to be solved before humanoid robots can scale. But a more fundamental question to ask is whether a bipedal robot is actually worth the trouble.
Dynamic balancing with legs would theoretically enable these robots to navigate complex environments like a human. Yet demo videos show these humanoid robots as either mostly stationary or repetitively moving short distances over flat floors. The promise is that what we’re seeing now is just the first step toward humanlike mobility. But in the short to medium term, there are much more reliable, efficient, and cost-effective platforms that can take over in these situations: robots with arms, but with wheels instead of legs.
Safe and reliable humanoid robots have the potential to revolutionize the labor market at some point in the future. But potential is just that, and despite the humanoid enthusiasm, we have to be realistic about what it will take to turn potential into reality.
This article appears in the October 2025 print issue as “Why Humanoid Robots Aren’t Scaling.”
Read the original article
Comments
Completely unsold on this take. The pace of development in China can't be ignored. The consumer market for a pretty dumb household chore bot is huge.
By datadrivenangel 2025-09-1117:521 reply I own at least three! Dishwasher, laundy washer, laundry dryer...
I would get a roomba but it can't do enough fine detail to be worth it.
By robots0only 2025-09-1117:242 reply and so is the safety margin for a humanoid. The consumer market is huge only if the robots are highly reliable and work very well both of which are not true at the moment. Things will change but it will take quite a bit of time and much more research.
Safety stops are much more challenging in a robot that tends to fall over without active balancing. Though with ISO 25785-1 in progress maybe there'll be a workable humanoid robot safety standard in a few years.
By omnicognate 2025-09-1118:49 Interesting, hadn't heard of ISO 25785-1. It appears to be for industrial robots, though, and explicitly excludes "mobile robots intended for consumer or household use".
By jpace121 2025-09-1117:35 Is the demand there at the price point and reliability levels that are currently possible?
By numpad0 2025-09-1117:46 They're all dog bots adapted to stand on back legs. None are true humanoids in the first place, none of them even have articulating pelvis.
By diamond559 2025-09-1123:41 Computers beat Gary Kasparov in chess in the 90s and we're all excited now bc it can write passable high school essays 30 years later. This is the real pace of "AI" and robotics development, decades not years.
Look at it this way. Useful Humanoid robots are at least as hard as useful self-driving cars. It took about 20 years to get from the DARPA Grand Challenge (can drive OK on an empty road) to Waymo (take one across town today.).
By AndrewKemendo 2025-09-1117:181 reply It’s actually going faster because it does not require public approval the same way that driving cars do because you’re in public space.
We’re seeing a lot of robotic trials happening in private warehouses and on private test ranges at pretty rapid scale
Beyond that the methods for transfer learning behavior cloning behavior authoring are very robust so that I can get joint angles directly from a human via instrumentation through vision or even commodity sensors which captured trajectories that can be immediately applied to robotic joint positions.
The real challenge is actually capturing demonstration recordings from humans because it’s the hardest thing to instrument. The core task is instrumenting data capture of existing human tasks that are not done through machines, such that they can transfer to machines.
This is easiest done with existing human operated robots because the instrumentation is free, so data can go directly into real2sim2real pipelines.
There might seem counterintuitive but most of the actual technical bits and bites are already there it’s re-orienting the economic and logistical process of labor execution that is the major challenge.
I will say though, I’m seeing less and less barriers there as time goes on. Employers really want to not have to hand human employees
By SequoiaHope 2025-09-1118:37 Municipal approval was probably only a minor part of the development time tho no? Google was operating Waymo in Mountain View for years before they expanded. The vast majority of the time cost was in development.
I think technology development can be faster thanks to better AI systems like VLA models, but I do think the time to real deployment will be long.
My pet issue is that the dexterity of the hands is still really poor. A human hand is incredible with what it can do.
I think between the general manipulation tasks, world understanding, and more these systems are still a long ways out for widespread use, though I wouldn’t be surprised if they find niche uses near term.
By modeless 2025-09-1117:54 ASIMO is 25 years old. PETMAN is 16 years old and Atlas is 12.
I think some of the car role will port to robot role.
Vision for cars already includes object detection, and the better that is, the better robot object detection gets. The same for "human ran out on road" would work for "walking in house, small human is now in front of me, stop!".
I wonder how much of the one will port to the other. A house has paths aka "roads", inside and out. Places the robot may walk, and not. So path navigation is a thing too. Maybe 'getting around' is mostly solved, while of course other challenges are still there.
Sort of replying to others in this part, the reason people are all hung up on humanform, is that our entire world is made for humans. Whether stairs, doors, sidewalks, doorknobs, cupboards, or even space to walk in a small kitchen... it's all made to work with human shape and size.
(Yes, while there is wheelchair access mandated, that doesn't extend to the inside of every home, and all the spaces in homes, and even then everything we have is designed to be operated by fingers/arms/hands.)
So if you solve humanform, the robot can go anywhere and manipulate/do anything a human can. That means no change to the environment when you get one. Right or wrong, that's why everyone is after humanform.
Tesla went into humanoid robots because they noticed what kind of thing was their AI architecture developing into.
They realized just how much of what an autonomous vehicle needs to do to navigate real world roads is similar to what an advanced robot would need to do to operate in real world environments. If they could get anywhere close to solving FSD, it would be an "in" on advanced robotics too.
The triumph of LLMs then made it glaringly obvious that the kind of advanced decision-making that you would need to power truly universal robots is no longer in the realm of science fiction, so a lot more companies followed.
> Tesla went into humanoid robots because they noticed what kind of thing their AI architecture was developing into.
Tesla went into humanoid robots to pump the stock. Musk recently claimed that 80% of Tesla's value is in Optimus.[1]
(What is it with US auto company management? Tesla did well for a while, then Musk got distracted. No new models in years, and the Cybertruck turned out to be a dud. Stellantis has cut Chrysler down to one line of mini-vans, raised prices on everything, and messed up Jeep, which killed sales. The dealers demanded the CEO be fired, which was done. GM is being GM, plugging along. The CEO of Ford seems to have a clue. He got a BYD car to drive around and has been telling everybody that Ford has to get that good and that cheap, and fast, which they are trying to do.)
[1] https://www.msn.com/en-us/news/technology/the-story-of-optim...
Tesla went into humanoid robots to pump the stock.
Everything else in your comment is more locked to direct observable info.
However this really isn't. And frankly every CEO wants a healthy stock price, that doesn't mean the primary goal is "pump the stock", which is what you are stating is the goal here.
That said, the Chrysler/Stellantis thing is just weirdosville. One thing I liked was this new platform:
https://www.caranddriver.com/ram/1500-ramcharger
but:
https://www.caranddriver.com/news/a64781518/ram-ramcharger-1...
They put all that work into it, hype, pre-launch, and it's even built on a common platform. The truck is basically the same as their pure-electric version, as the engine is simply used as a generator to charge. No major changes to the rest of the vehicle, just an added engine.
You may wonder why I like this? Well I was hoping it's carry over to smaller Jeeps or even cars. I live in a rural area, and it gets quite cold here (-40C sometimes), which really reduces range. I want a PHEV, but also with a backup power plant.
This is that.
It also reduces a lot of the complexity of a hybrid. No dual drive train, or one part driven by gas engine, the other electric.
I can see myself driving much of the summer with the engine only coming on rarely, but then needing the engine more and more in the winter. But at least I'd be on battery most of the time, even then.
Heck, I can drive 100s of km without there being a fast charger here.
We need bridge vehicles for the time being.
But my rambling really goes back to, they developed the platform, started selling it, but held off on this final piece because of.. reasons?
Weirdos.
> Tesla went into humanoid robots to pump the stock. Everything else in your comment is more locked to direct observable info. However this really isn't.
Musk says it is: "Tesla CEO Elon Musk said on X that about 80% of his automaker’s value will come from Optimus humanoid robots. Musk said in mid-2024 that Optimus robots could make Tesla a $25 trillion company, equal to more than half the value of the S&P 500 at the time of his comment."[1]
I'd like a good $35,000 electric Jeep myself.
[1] https://www.cnbc.com/2025/09/02/musk-tesla-value-optimus-rob...
By b112 2025-09-138:54 Sure, but that's discussing the outcome of a goal. That doesn't make it the goal. To highlight this, they've been working on Optimus for what? A decade?
'Pump the stock' has more of a bait and switch, or a scammish connotation, at least to me. Every company must be aware of stock price to a degree, and tout their successes and products and goals. That's not pumping, which is often conjoined with pump'n'dump.
But anyhow.
What I like about that platform, is you could have that $35k electric Jeep, and I could have a $45k electric Jeep with built in generator and other fluff modz, but it would be built on the exact same platform. Same parts except for the generator+tank.
One of the benefits could be space for extra bling. You could put a water tank on the mounting brackets for the fuel tank, or something else interesting. You could convert your model to a gen model, even buying parts from the wreckers on the cheap, should your circumstances change. Or the reverse! Remove the gen+tank if desired.
I can imagine needed the gen+tank for 4 years, then maybe I suddenly have sufficient fast chargers locally...
Heck selling a model with the wiring + tank would add minimal weight. Even just the wiring. Imagine being able to rent an engine for a week, a month, a year? The mount rails and a generic, standardized connector is there, the software supports it, an open API.
So much could be done, but of course won't.
But I do, as maybe you can tell, really like the idea of a shared platform.
By numpad0 2025-09-125:41 Elon wanted to automate final assembly. Final assembly in cars is anything but final; it's most of what regular people consider "building" phase of the car where the body on a cart rolls forward and human workers attach parts to it until it rolls off at the end.
A lot of final assembly processes are already highly automated, albeit with human hands as tip of robots. Heavy objects such as entire assembled dashboard spanning an entire width of the car, or seats complete with headrest, are slid in from side riding on a carriage, often following a human hand that route cables and to confirm alignments for attachment.
Joint wears of worker humans embedded in the processes as robots are tracked and managed as well, by computer simulation, data tracking, and optimization.
Tool used, such as drills, are highly computerized as well, automatically setting and logging turns and torques, types and numbers of screws removed from the workstation shelf, etc., but they are still manipulated and articulaed by humans, as robots aren't good enough.
Robotizing final assembly is the holy grail for which Honda was doing robots since late 80s until Japanese economy collapsed and AI researches stopped. Steve Jobs was dreaming the same lights-out factory concept during NeXT years. More recently, Foxconn bought robots in hundreds at one point in hope that it will be useful for iPhone. Sony actually shipped some numbers of PS4 from a fully robotized prototype line.
The difference between them and Tesla is that they're so far the latest and about the least actually committed to it, obviously due to lack of engineering talent outside of UI software.
By HarHarVeryFunny 2025-09-1119:08 > Humanoid robots are at least as hard as useful self-driving cars
A self-driving car only has to do one thing - drive. It's also got a stable wheeled base and only a couple of degrees of freedom - got to steer and regulate it's speed.
Even if the only thing you wanted the humanoid robot to do is drive your car, it'd be massively harder for it since it's got all those degrees of freedom, will be bouncing around in the drivers seat, and presumably doesn't even know how to drive.
If the humanoid is more than a gimick - meant to be general purpose, then it needs an AGI brain and ability to learn for itself. It's not going to be learning in a simulator like your FSD car - it's be learning on the road like your teenage kid.
(a few specific towns, which have done various things to smooth operating conditions for Waymo.)
It's useful, don't get me wrong, but when Waymo can handle Cairo and Rome, I'll consider it a solved problem.
By AlotOfReading 2025-09-1117:562 reply What specific things do you think the "towns" of SF, LA, NYC, and Tokyo have done to smooth operating conditions for Waymo?
Drive according to western standards. If you've never driven outside of those, then it's hard to imagine. No one follows the laws. Lanes are entirely made up, no one follows the lines on the road. If you're not agressive to the point of someone in LA about to shoot you, you won't get in.
I ride Waymos and I believe they can be made to work in Rome et al, but honestly, I doubt most human drivers can drive in eg Mumbai.
By AlotOfReading 2025-09-1222:48 I'm familiar with Indian traffic. Those events are more frequent in some places, but they happen everywhere. In any case, it's not what the OP is talking about. They were talking about cities improving conditions to make things easier for AVs. LA wasn't Mumbai before Waymo and it still isn't after.
By stackedinserter 2025-09-1117:501 reply Maybe cities like Cairo are problems, not the algorithms that can't drive there.
The problem with respect to what? The end-goal of self-driving cars (and humanoid robots) is to work in the environments created for humans. Otherwise we can just put down rails across all cities and call it a tram, or design purpose-built robots for all tasks.
Edit: Stated more explicitly: the human world is the way it is because of many reasons and can't always be changed naively (it's not like nobody in Cairo has thought about improving the traffic situation, or architects haven't thought about the ease of cleaning different flooring material). Robots which are general purpose with respect to their human-like capabilities must necessarily also accept a world in which humans live.
By stackedinserter 2025-09-1119:10 The end goal of is to make them work in reasonable environments. If it works fine in 90% of cities but doesn't work in Cairo, then fuck Cairo, no driverless cars for them.
We have Roombas. I saw a number of Husqvarna lawn robots in Sweden (I've seen none in the U.S. so far). But neither of these are exactly flying off the shelves.
Humanoid robots feel like they're decades away for being something people would want.
I've owned several Roomba type robots, both actual Roombas and competing brands. None of them have really saved any time or labor. They always get stuck under furniture or tangled on charger cables. They don't work on stairs. And clearing dog hair from the roller is a huge hassle. I fully expect to still be paying a human cleaning service decades from now.
> And clearing dog hair from the roller is a huge hassle.
I'd like to get a robovacuum/vacuubot but I'd assume that cat hair fouls things up just as much as dog hair does.
By sonofhans 2025-09-1120:35 Not really. A golden retriever has hair longer than any cat does, and of course large dogs shed much more.
By ajdude 2025-09-1118:37 > I fully expect to still be paying a human cleaning service decades from now.
I'm curious if a cleaning service wasn't an option, would the Roomba be worth the saved time compared to doing the cleaning yourself?
I also own a Roomba, but I don't have a cleaning service so my options are either do 100% of the cleaning or let the Roomba do its thing and manually take care of the difference.
For me it's just one less room I have to sweep.
By stackedinserter 2025-09-1117:45 Of course there always be "There is no reason anyone would want a computer in their home" people, that's life.
Roomba is pretty mediocre at a single job it's kind of able to do.
Humanoid robots _potentially_, _hypothetically_ can do anything that human can do because they are designed for environment, tools and equipment that we designed for our bodies.
> But neither of these are exactly flying off the shelves.
Roombas and lawn robots are all extremely popular.
By jsbisviewtiful 2025-09-1117:46 Robot vacuums are popular but maybe they were referring to the iRobot brand, which is rapidly failing.
By fluidcruft 2025-09-1117:40 I have vacuum robots. I'm considering a lawn robot. My suburban city has two large ones mowing the parks, had never heard of them previously. Mostly worried about pets and critters.
