Every new computing era begins with a question: who needs this?
When mobile phones first appeared, it was easy to treat them as an expensive convenience for a narrow group of users. Similar skepticism now surrounds humanoids, quadrupeds and other Physical AI systems: why does a home, warehouse, hospital or factory need one? FlyPig AI believes that is the wrong time horizon. The more useful question is what happens when hardware cost, reliability, software abstraction and developer access improve together for ten years. That is the pattern that turned mobile computing from a device category into an application economy.
The smartphone lesson is not the phone. It is the platform shift.
Android created an openly available operating-system base that device makers could customize, while chipset and reference-design vendors reduced the amount of engineering required to produce a competent handset. MediaTek's public reference-design history shows how hardware, software support and pre-integrated capabilities can shorten time to market for OEMs. The result was not one standardized phone. It was an explosion of differentiated products built on increasingly standardized foundations.
Physical AI is beginning to expose the same pattern
Robot builders can increasingly start from complete bodies, compute modules, motion controllers, perception stacks and open robotics software rather than from raw motors and custom boards. XGO demonstrates the small-scale developer version. Unitree demonstrates larger programmable quadruped and humanoid platforms. NVIDIA's 2026 Isaac GR00T Reference Humanoid Robot makes the analogy much harder to dismiss: NVIDIA explicitly combined a Unitree body, dexterous hands, Jetson Thor compute and an open GR00T software stack as a reference design for humanoid research.
Hardware can be valuable without being the final value pool
FlyPig AI does not argue that hardware companies cannot make money. Some will. The stronger historical warning is that capital-intensive hardware categories often face brutal competition once manufacturing knowledge spreads and specifications converge. Digital cameras, LCD panels and many consumer-electronics categories show how quickly technically difficult hardware can become a scale game. In that environment, the highest-leverage businesses frequently emerge above the component layer: operating systems, developer ecosystems, applications, services and proprietary user or operating data.
The phone analogy has a hard limit
Robots act in a physical world that is far less standardized than a phone screen. Payload, reach, terrain, safety, manipulation, battery life and duty cycle differ by use case. Physical AI is therefore unlikely to converge on one universal body. Modularization will happen by robot class and capability envelope. That makes the software abstraction above the body even more strategically important.