Summary: RoboStrategy, a closed-end fund focused on robotics, surveyed a US market with over a hundred humanoid robot companies — and its CEO Andrew Kang named only Tesla, Figure AI, and Apptronik as picks to build real consumer products. His reasoning is not about capital but about a global scarcity of top robotics talent. For builders, it’s a useful lens on why crowded hardware markets still concentrate into few winners.
Why only three winners out of 112
Per a TechOrange report citing Kang, the US has 112 companies working on humanoid robots — manufacturing, components, software, and AI. China has over a hundred similar firms, and European startups keep appearing. Yet Kang expects the industry to produce few winners, not dozens.
His argument: the bottleneck is expertise, not funding. A top robotics company needs specialists in robot learning, robot behavior, computer vision, actuator engineering, battery engineering, and product management. By his estimate, only a few hundred people worldwide hold genuinely top-tier skills across those areas — and gathering them under one roof while raising billions is genuinely hard. Software scales; elite robotics engineers do not.
This is a familiar pattern to anyone tracking Skild AI’s push to build a universal robot foundation model — the scarce asset in physical AI is the team that can turn models into machines that actually work in the real world.
The Optimus question
Tesla is the most contested of Kang’s three picks. Publicly shown Optimus demos lag Figure AI and Apptronik, and even trail Agility’s Digit or 1X’s NEO. Kang doesn’t dispute this — he pointed to Figure’s live package-picking demo of the Figure 03 as evidence that Figure leads Tesla on AI today. His explanation is simpler: Elon Musk is stretched across too many companies. Kang’s bet is that Musk delivers late but delivers, based on his track record of eventually shipping things others couldn’t.
Notably, Kang’s conviction doesn’t fully extend to his portfolio. Apptronik holds 14.2% and Figure AI 13.8% of RoboStrategy’s fund; Tesla isn’t in it at all, and Tesla stock is down nearly 20% this year. Talk and positions diverge — worth remembering when evaluating any fund’s public thesis.
The quiet industrial tilt
The largest single holding is not a humanoid company at all. Standard Bots, focused on industrial automation, makes up 31.4% of the fund. Add positions like Dyna, Dexmate, Maven Robotics, Path Robotics, and Eccentric Machines, and roughly a third of RoboStrategy’s assets sit in traditional industrial robotics. Near-term cash flow lives in factories, not households — a practical hedge on the humanoid timeline.
The cost math and the infrastructure caveat
Kang sizes the robot market in the tens of trillions of dollars. His back-of-envelope: depreciate a humanoid over seven years, run it three shifts, and its annual labor cost lands around $5,000–$7,000 for output equivalent to roughly three human workers. At that price, many jobs that never penciled out suddenly do — which he ties to the US reindustrialization push and the FCC’s move to put foreign-made robots on a controlled list.
He predicts millions of humanoids shipping between 2028 and 2029, and homes adopting them widely by 2030. The caveat he flags matters most for builders: unlike spinning up a million chatbots, physical robots require actual manufacturing capacity, and infrastructure has to arrive first. Plan hardware-adjacent roadmaps with that lead time in mind — the demand curve may show up faster than the supply chain can.