Humanoid robots have become the mascot of Silicon Valley’s physical AI boom.
They dance in viral videos, bang out tunes on the piano, and are headed to Wall Street as Agility Robotics plans to go public. They promise a new workforce that never sleeps, never ages, and can work in spaces built for people.
But as AI moves off screens and into machines, some Silicon Valley investors are not buying the humanoid hype.
“The theory around humanoids is that the world is designed for humans, so let’s create an embodiment that looks like a human,” said Ajay Agarwal, a partner at Bain Capital Ventures and an early backer of Kiva Systems, the warehouse robotics company Amazon bought for $775 million in 2012. But Agarwal believes that humanoids could prove to be a “parlor trick” with few practical uses.
His concern reflects growing skepticism among some investors that robots need to mimic the human body at all. In their view, humanoid companies inherit the constraints of the human form instead of designing the machine best suited to the job. Legs, for example, have to support a heavy torso battery, which means the robots consume more power and create the safety risk of falling over.
“There’s a reason why humans fly planes and drive in cars,” Agarwal said. “Because wheels and wings are more efficient than walking.”
‘The humanoid fallacy’
A slew of big-name investors are steering clear of humanoids in favor of other robot forms. Khosla Ventures and former Google CEO Eric Schmidt have backed Genesis AI, which unveiled a wheeled general-purpose robot with no head or legs last month. Bain Capital Ventures and Sarah Guo’s Conviction have invested in Sunday Robotics, which is building a wheeled home robot. Neil Mehta’s Greenoaks has funded The Bot Company, a wheeled home-robot startup founded by former Cruise CEO Kyle Vogt.
Eclipse, a firm founded in 2015 to back startups building in the physical world, has an extensive robotics portfolio that notably does not include humanoids. Instead, the firm has backed autonomous vehicles and robots for warehouses, construction, healthcare, and retail.
Jiten Behl, a former Rivian executive and partner at Eclipse, said copying the human form often makes a robot less useful.
“For the vast majority of tasks that happen inside a manufacturing site, you don’t need to walk, and you don’t need to stand,” said Behl, an investor and board member at Mind Robotics, a Rivian spinout building AI-powered industrial robots. “The smartest attempts are going to be those that look at the use case and fit the right form factor to the use case instead of trying to average it out to a humanoid form factor.”
Ghazwa Khalatbari, an investor at Creandum, reached a similar conclusion after moving from New York to San Francisco last year and noticing that humanoids were the hot topic among investors.
“But then I watched some of the demos and thought, ‘This robot is putting a plate in the dishwasher slower than my grandmother,'” Khalatbari said.
That disconnect inspired a research project arguing against the “humanoid fallacy”: the assumption that because the world is built for humans, robots must also look human. Khalatbari and Lux partner Deena Shakir argued that robot bodies should be designed around the job they are meant to do. Physical Intelligence, a Lux portfolio company, is building AI models designed to work across different types of robotic bodies.
“I think humanoids will be one species in a much larger ecosystem,” Khalatbari said. “In a hospital 20 years from now, you might have a humanoid helping a patient out of bed, an autonomous cart delivering medication, a pair of dexterous hands sterilizing surgical equipment, and a Roomba cleaning the floors.”
‘The humanoid economy’
There are real reasons startups and corporate giants are betting heavily on humanoids.
If humanoids work, the prize is the entire physical economy. It would be a world where “human labor is optional,” said Jonathan Hurst, cofounder and chief robot officer at Agility Robotics. “Humans will have this tireless partner to amplify our ambitions and goals.”
Last year, humanoid robot companies raised more than $6 billion, up more than 300% from the 2024 total, according to PitchBook data. A forecast from Morgan Stanley predicts the market for humanoid robots will hit $5 trillion by 2050, with a billion units in circulation.
The case for humanoids starts with the fact that the world is built for people. Elon Musk has defended the shape, arguing that the humanoid form makes sense if the goal is a robot that can do everything humans can do.
“It turns out humans evolved to the shape and capabilities that we have for good reasons,” Musk said at the All-In Summit last year. “There’s value to having four fingers and a thumb.”
Now companies are trying to prove the humanoid bet can become a business. Hyundai-owned Boston Dynamics is working to deploy Atlas in factories in 2028. Zachary Jackowski, the engineer who leads the Atlas program, said the goal is a “software-defined factory,” where humanoids can be assigned different tasks as needed.
Agility has seen early traction. Digit, its humanoid robot, is deployed across nine customer facilities, including Amazon, Toyota, and logistics company GXO. The company said it has more than $300 million in multiyear orders for the next generation of Digit.
San Jose-based Figure AI, most recently valued at $39 billion, is starting deployments in logistics and distribution centers this year. Palo Alto-based 1X plans to ship more than 10,000 home humanoids later this year.
Jeff Cardenas, CEO of Texas-based humanoid startup Apptronik, said he sees the biggest long-term potential in legged humanoids because they could eventually do any task a human can physically do. But the company also has a wheeled version of its humanoid, which Cardenas expects to be deployed sooner because it poses fewer safety challenges.
Modar Alaoui, general partner at ALM Ventures and founder of a humanoid-focused conference, is betting on what he calls “the humanoid economy.” He expects humanoids to arrive in waves: first in industrial settings such as manufacturing and warehousing, then in consumer-facing roles at theme parks, shopping centers, and airports, and eventually in homes.
China is already pushing in that direction. The government is launching a nationwide training program to help commercialize humanoids, giving local governments and state-owned enterprises less than six months to prove the technology can work in production and service settings.
Chinese companies already dominate the humanoid market. Firms such as Unitree and UBTech accounted for about 90% of humanoid robot shipments last year, according to Omdia. Unitree is preparing to go public and is targeting a valuation of up to $7 billion.
Still, even some humanoid believers draw a line between useful humanoids and robots that merely look human. Hurst, a robotics professor at Oregon State University who founded Agility in 2015, said he is wary of companies blindly copying the human form.
“They’re building machines that look like a person, but really are just animatronics,” Hurst said of some humanoid startups. “When we end up with an actual configuration that is humanoid, we’ve arrived at that through 15 to 20 years of function-first, physics-first research, understanding exactly why it looks the way it does.”
Digit, for example, departs from the human form with backward-bending legs that let it reach from the floor to overhead shelves without its knees hitting warehouse racks.
In Hurst’s view, humanoids are inevitable because there will always be a need for robots in human spaces. He agrees with skeptics that specialized robots will be far more common, but believes humanoids will still have a lasting place.
“Purpose-built automation is going to outnumber humanoids by quite a lot,” he said. “And there will also be millions and millions of humanoids, billions perhaps.”
Read the full article here
