This Phantom humanoid robot by Foundation is working in a factory. Some of them, however, will be working in military applications if the company’s plans come to fruition.
John Koetsier
Most of the hundreds of humanoid robot startups around the world are building robots for work, whether in the factory, the community, or the home. Silicon Valley startup Foundation is too, but it’s also explicitly targeting the U.S. military as a customer for Phantom, its 5’9″, 180-pound humanoid robot. In fact, Phantom will both carry lethal weaponry and fight, and CEO Sankaet Pathak plans to build 50,000 of them in near-record time: by the end of 2027.
“You should really work hard to give the U.S. military smarter tools so that they can be more effective,” Pathak told me in a recent TechFirst interview. And he’s not shy about giving Phantom an M4 Carbine, either, adding that “if you’re first body in and you’re docile, then the enemies are not going to really expose themselves. So you have to be first body in and deadly.”
Of course, military use is not the only plan for Foundation.
The company has built Phantom in near-record time. Within 18 months of establishing Foundation, Pathak already had a production robot hard at work running actual production tasks at (undisclosed) partner locations. That’s almost as fast as Apptronik, thanks to two smart acquisitions in AI and next-generation actuators that vastly accelerated progress.
Foundation’s scaling plan is even more aggressive: 40 robots this year, 10,000 next year, and 40,000 humanoid robots off the production line in 2027. Pathak acknowledges that it’s extremely aggressive, but says there’s a “non-zero chance” the company can make it happen.
One of the reasons: the Foundation team, which includes top talent recruited from companies like Tesla, 1X, Boston Dynamics and SpaceX.
“Our head of manufacturing is an ex-Tesla manufacturing director,” Pathak told me. “He worked on the Model X and Y ramp. So the lessons are just baked in: don’t try to automate too quickly.”
If Foundation can manage the insane growth, Pathak is confident customers will appear. He’s intending to lease robots, not sell them, and he doesn’t need hundreds of customers either.
“We don’t need 50 deals,” he told me. “We need like five really high quality large deals, and they can scope out to be hundreds of millions of dollars in annual recurring revenue purchase orders.”
In fact, if Foundation hits its production goals and sales happen the way Pathak envisions, Foundation becomes an overnight unicorn, multiple times over: 50,000 robots built over 2026 and 2027 would lease out for about $5 billion in annual recurring revenue, he says.
A little bit of math tells you that means Foundation currently plans to lease robots for about $100,000/year each. That seems steep, giving restaurant, warehouse, or production line labor might cost an average of around $40,000 annually, but it’s actually a discount on wetware (humans like us) because the robots can work nearly 24/7, meaning each robot does the equivalent of about three to five humans. Assuming maintenance and repairs are included in the $100,000, that’s a pretty significant discount.
Cost versus return for robot labor compared to human labor
John Koetsier
At near-full utilization, robots could save employers $166,000 each year. With a more realistic calculation including both some downtime for robots and some human overseers, you’re looking at about $90,000 in annual savings.
This all assumes, of course, that the robot is both as fast as humans and as good as humans: something no humanoid robot manufacturer has yet been able to make happen. So the smart money here will add at least a few years to Foundation’s best-case napkin scratches, or discount the value of the robot labor significantly as the hardware and software stack ramp up to full human-equivalent capability.
And that’s in spite of Foundation’s high-end robot “muscles” that can operate for multiple shifts without overheating, using highly efficient, backdrivable actuators that make Phantom both powerful and comparatively safe around people.
But there’s also the military applications.
Like it or not – and anyone who’s watched the Terminator move probably doesn’t – humanoid robots have a huge amount of utility in military applications. They can be used to do grunt work that soldiers or support staff don’t have to, they can ferry weapons and ammunition to soldiers, they can take risks that no good officer wants his or her fighters to undergo. We’re already seeing similar uses of wheeled robots in the Russia-Ukraine conflict.
And yes, this would include carrying a weapon into a building in street fighting scenarios, or being the first one over a hilltop, or entering a cave to search for enemy combatants.
In fact, Pathak argues that humanoid robots could also make military force more precise, not less. Rather than using air strikes or heavy weapons, the military could use a ground-based humanoid robot that can enter buildings, assess situations directly and make the right call. While Phantom would not be fully autonomous in lethal decision making, Foundation envisions a model similar to current military drones, where robots handle movement and navigation themselves while humans back at base (or nearby but safely hidden) retain control over targeting. If that approach holds, humanoid robots like Phantom may ultimately change not only how wars are fought, but when they are avoided, by shifting deterrence toward visible, scalable robotic capability rather than human deployment.
And, rather than the cold-blooded killing image that Terminator-style humanoid drones might have, Pathak argues they would actually be highly effective peacekeepers.
“I do think if the U.S. military had like 100,000 robots that they could just demonstrate work, it would by and large end a lot of wars before they start,” he says.”
That makes sense: the U.S. would be more likely to send them in than humans who might get killed and cause political backlash. Knowing that, rebels, insurgents, terrorists, or other enemy combatants might be less likely to start wars in the first place.
But the opposite might be true as well: the military would be so much more likely to use fighting humanoid drones that wars could be come more likely.
Either way, the ethics of humanoid robots is about to get even more complicated. And if you think that China or Russia isn’t thinking of building warbots, think again. Essentially, though not in humanoid form, they already are. So is the U.S. military, which has had Predator drones for three decades now.
And the first automated and uncrewed weapons of war might date back to World War II: Germany’s V-2 rockets. They had a very crude and rudimentary version of autonomous guidance and control, making them perhaps the first war robots ever created.
Now, we’re just upping the level of sophistication. And, of course, lethality.
Source: https://www.forbes.com/sites/johnkoetsier/2025/12/16/humanoid-robots-for-war-and-work-startup-plans-to-build-50000-by-end-of-2027/
