Why Software Founders Should Pay Attention to the Humanoid Robot Race (Even If They’ll Never Build One)

In 2007, a lot of people looked at the original iPhone and concluded it was a hardware story. A new device. A new interface. Interesting, sure. But hardware was Apple’s thing, and most software builders felt comfortable watching from the sideline.

What happened next was not a hardware revolution. It was a software platform revolution that happened to arrive inside a device. The App Store opened in 2008, and within a few years, entire industries had been rebuilt from the app layer up. Maps. Taxis. Food delivery. Banking. Photography. Not one of those disruptions required anyone to manufacture a new phone.

I keep thinking about that moment as I watch Meta acquire robotics startup ARI, Amazon gobble up humanoid robotics companies, and every major tech player race to put a bipedal machine in a warehouse or a home. The coverage frames this as a robot race. A manufacturing competition. A bet on which company’s exoskeleton will win. And most software founders I know are watching with mild interest and no urgency, because it reads like a hardware story.

It is not a hardware story.

Humanoid robots are a new platform. And new platforms always do the same thing: they create an application layer that nobody owns yet, right at the moment when being first matters most.

Think about what happens when there are millions of robots operating in homes, warehouses, retail floors, and care facilities. They will need software to understand context, to communicate with humans, to escalate problems they cannot solve autonomously, to be remotely diagnosed when something goes wrong. The robot manufacturers will be good at the robot. They will not be good at the vertical software stacks that make robots useful in specific industries. That gap is where founders live.

Consider the unsexy details that actually determine whether a robot deployment succeeds or fails. Does the robot know when to ask for help? Does it know who to ask? Can a remote technician see what it sees and walk it through an exception? Can a manager review what the robot did during a shift the same way they would review a human employee? These are software problems. Workflow problems. Communication problems. None of them require building a robot. All of them require someone to build the layer that sits on top of one.

There is a version of this that applies to almost every vertical. Fleet operators need logistics software that accounts for robot downtime differently than human downtime. Care facilities need compliance and documentation tools that work whether the caregiver is human or machine. Retailers need planogram auditing tools that plug into a robot’s camera feed rather than a phone. Training and onboarding platforms need to be rebuilt for a workforce that includes both people and robots working side by side.

None of these are science fiction. They are the ordinary, unglamorous software infrastructure problems that will exist the day after the first big robot deployment in each industry. The same way that the existence of GPS-enabled smartphones created an obvious infrastructure layer that Uber and Lyft and DoorDash and Google Maps each occupied, the existence of embodied AI in physical environments creates a new infrastructure layer that is currently unoccupied.

The founders who will own embodied AI software opportunities are not the ones building the robots. They are the ones thinking right now about what becomes possible, necessary, or obvious the moment the robots exist at scale. That thinking has to happen before the deployment wave arrives, because by the time it is obvious to everyone, the platform winners will already have market position.

This is not a prediction that software founders should pivot to robotics. The opposite, actually. The hardware race is well-funded, well-covered, and crowded with well-capitalized competitors. The software layer is open. It requires domain knowledge about industries, not about servos and actuators. It requires understanding how humans and machines need to communicate, which is something that software founders already know how to think about.

The question worth sitting with is not whether humanoid robots will work. They will, in some form, in some industries, faster than the skeptics expect. The question is: what is the “Uber moment” that only becomes possible when there are millions of robots in homes and warehouses, and are you thinking about it yet?

The window for being early is now. Not because the robots are here. Because platforms always reward the people who show up before they are.