Humanoid robots used to live in science fiction. Now they live in VC portfolios, and the money flowing into them is growing quickly.
The idea of robots that look, move and operate like us has captured imaginations for years, dating back to Leonardo da Vinci’s 1495 designs for a mechanical knight. Today’s robots have come a long way: they do backflips, fold laundry, and even hold a conversation.
As humanoids become more of a reality, investment has flooded into the space. More than $5 billion has been invested in humanoid startups this year, according to PitchBook data, surpassing last year’s record total.
With 1 in every 4 dollars invested in robotics going to humanoids, deals in the hundreds of millions have become the norm. China’s Robotera raised more than $200 million just over a week ago, while University of Texas spinout Apptronik pulled in $520 million earlier in the year, tripling its valuation in 12 months.
Behind the buzz, though, questions about use cases, deployment and the path to returns are proving harder to answer than the fundraising rounds might suggest.
“There’s no strong, rational reason [behind current investment levels],” Fady Saad, GP at robotics-focused VC firm Cybernetix Ventures, said. “They’re cool, but there’s a psychological attachment or obsession with them that just isn’t practical.”
So why are checks being written?
A ‘limitless’ market
Beyond the novelty of humanoids, economic factors are also spurring investment, according to Sam Baker, a deep-tech investor at Berlin-based VC firm Planet A.
According to Goldman Sachs research, the cost of manufacturing a unit fell 40% from 2023 to 2024, well ahead of the 15%-20% decline expected. Cheaper components, such as motors and sensors, along with improvements in design and manufacturing processes, have brought costs down to a level an enterprise customer could reasonably expect to pay.
Cheaper humanoids mean more customers and greater penetration, especially when they arrive in homes. Morgan Stanley estimates that a humanoid model could cost as little as $15,000 by 2050, with about 10% of US households having one by then.
Furthermore, most countries face shortages in agricultural, retail and manufacturing workforces. Humanoids could, in theory, fill the gap if they become more accessible.
We’re in a classic hype cycle, but what isn’t overhyped is the fundamentals of the technology
The second driving factor, Baker says, is the economic potential of a robot with no fixed purpose and, in theory, no ceiling on what it can do.
“There’s a versatility aspect to humanoids,” Baker said. “With specialized robotics, the economic value potential is constrained by whatever market they’re sold into and designed for. But there’s this general impression that you could have a humanoid that fixes your laptop, cooks you dinner and builds you a car, all in just one machine. From an investor’s perspective, there are almost no limits to the addressable market.”
Investment bank RBC Capital Markets projects that the global humanoid robot market could reach $9 trillion by 2050, driven by the range of potential use cases.
The problem with legs
Versatility is a key factor in the investment case for humanoids, but Saad believes it is precisely what diminishes their value in a factory or warehouse environment.
“If you’re on a manufacturing floor, most facilities require 99.99% accuracy, which is hard to achieve,” he said. “General-purpose robots simply cannot do that, because they’re not optimized for the specific use cases. They could be applied to use cases that don’t require that level of precision, like moving something from one place to another, but why do you need such a complex and expensive machine for such simple tasks?”
While there is strong demand for humanoid developers, Saad pointed to a lack of customer contracts as a sign that capital is racing ahead of the technology. A few car manufacturers, including Schaeffler and BMW, have signed partnerships to deploy these machines. Still, Saad said that the deals he has seen so far reflect more of a desire to appear innovative than a serious appetite for integration.
According to Baker, humanoids suffer from a different flaw, one that goes to the heart of what makes them humanoid in the first place: their form.
“You don’t need legs,” Baker said. “You often hear the argument [in favor of humanoids] that factories are designed for humans, so human forms are best, but that’s not strictly true. If you want to speak to an engineer in the office or have lunch in the canteen, sure, but on the shop floor, there are forklifts, pallet jacks and all kinds of machines that move things around without legs.”
From a physics perspective, legs are something of a hindrance for robots. Stability and balance are inherently more difficult due to their high center of gravity, especially when carrying loads or moving quickly. Legs require more joints, actuators and sensors than wheels, which means more points of failure and higher costs.
Some startups are developing legless humanoids, such as Boardwalk Robotics, whose robot Alex is an upper torso, but they are in the minority, and the major players in the space are focused on bipedal forms.
Home, sweet home?
While Humanoids’ usefulness in a factory setting is questionable, the real holy grail for startups is consumer adoption. Leading startups are already heading in that direction. Figure AI‘s newest model is focused on households and handles chores such as laundry and dishwashing. 1X‘s NEO robot, designed for domestic spaces, is scheduled to ship this year.
But if creating a robot that can autonomously move around a warehouse is tricky, making one for the home, with children, pets and piles of laundry in unexpected places, is orders of magnitude more complex.
Latest Research
Too Central to Fail: The AI Venture Ecosystem Built Around One Company’s Strategy
“We’re a long way from humanoids in homes,” David Ripert, partner at Finnish deep-tech VC FOV Ventures, said. “It’s a lot easier for a robot to understand how to think and act in an environment like manufacturing, because it’s very structured and predictable. There’s not enough training data for households, and a lot of them are being operated remotely by humans.”
Even if humanoids can function in homes, the safety risks are likely to keep them from consumers, according to Saad. Official standards are still in development, and unified regulations are not expected for several years.
The physical concerns are overshadowed by the risk that a machine with cameras and an internet connection in your home could be hacked.
Functions for the form
Despite the myriad of criticisms facing humanoids, even skeptics agree that there are use cases for the form.
Planet A’s Baker believes that these robots may not live in people’s homes, at least in the short term, but they can serve consumers in another way: through retail.
Similar to a manufacturing environment, retail is structured and already being automated, making deployment easier than in households. Additionally, stocking shelves, carrying items and greeting customers have a much higher tolerance for error than assembling car parts or other manufacturing tasks. The lower precision requirement means the technology doesn’t need to be as mature to be commercially viable.
According to Cybernetix’s Saad, humanoids can serve as replacements for humans in tasks where their lives could be at risk. Space is an area of interest, as shuttles and rockets are designed specifically for humans. It currently costs about $130,000 per hour to send an astronaut to space. Meanwhile, an industrial-grade humanoid would cost an estimated $250,000 to buy outright, meaning it would pay for itself in under two hours.
But these more limited use cases are a far cry from the vast opportunity investors are currently betting on. There is, however, a silver lining. Whatever shape the technology ultimately takes, the underlying technology remains valuable.
“When you look at technologies that have ended up changing the world, the journey to mass propagation tends to be unstable,” Baker said. “We’re in a classic hype cycle [with humanoids], but what isn’t overhyped is the fundamentals of the technology. Just because it is being developed for a bipedal machine doesn’t mean that’s the ultimate form it will take.”
Recent News
Weil Gotshal quietly incubated an AI-native Big Law challenger before its collapse
Why Wall Street wants to trade AI compute like oil
Learn about our editorial standards.