Physical AI and robotics now make up a meaningful chunk of my portfolio. Still feels like the market hasn't caught up to how big this shift actually is.
Most people are still anchored to software AI — chatbots, image generators, coding assistants. That's already a massive wave. But physical AI is different. We're talking about intelligence that can manipulate the real world: robots that pick, pack, assemble, cook, clean, drive, build.
The unlock isn't just better algorithms. It's the convergence of three things:
1. Foundation models that can generalize across tasks (not just narrow robotics code)
2. Sim-to-real transfer getting good enough that you can train in simulation and deploy in messy real environments
3. Hardware costs dropping while capability density goes up — compute, sensors, actuators all improving fast
What makes this undervalued? A few things:
First, people underestimate how fast robotics companies can scale once the software works. Unlike pure software, there's a hardware moat and real-world deployment complexity, but once you crack it, the TAM is enormous. Every warehouse, factory, kitchen, construction site, farm.
Second, the market still treats robotics like a niche vertical. But if AI is eating software, physical AI is eating labor. That's a way bigger market.
Third, most investors got burned by robotics hype cycles in the past — Boston Dynamics doing backflips but no business model, or clunky industrial robots that needed armies of engineers to deploy. This time it's different because the AI layer is general-purpose and adaptive.
I'm seeing early traction in logistics, manufacturing, food prep, and elder care. The companies that figure out the go-to-market and can actually deploy at scale are going to be massive. This isn't a 5-year horizon anymore. It's happening now.
Most people are still anchored to software AI — chatbots, image generators, coding assistants. That's already a massive wave. But physical AI is different. We're talking about intelligence that can manipulate the real world: robots that pick, pack, assemble, cook, clean, drive, build.
The unlock isn't just better algorithms. It's the convergence of three things:
1. Foundation models that can generalize across tasks (not just narrow robotics code)
2. Sim-to-real transfer getting good enough that you can train in simulation and deploy in messy real environments
3. Hardware costs dropping while capability density goes up — compute, sensors, actuators all improving fast
What makes this undervalued? A few things:
First, people underestimate how fast robotics companies can scale once the software works. Unlike pure software, there's a hardware moat and real-world deployment complexity, but once you crack it, the TAM is enormous. Every warehouse, factory, kitchen, construction site, farm.
Second, the market still treats robotics like a niche vertical. But if AI is eating software, physical AI is eating labor. That's a way bigger market.
Third, most investors got burned by robotics hype cycles in the past — Boston Dynamics doing backflips but no business model, or clunky industrial robots that needed armies of engineers to deploy. This time it's different because the AI layer is general-purpose and adaptive.
I'm seeing early traction in logistics, manufacturing, food prep, and elder care. The companies that figure out the go-to-market and can actually deploy at scale are going to be massive. This isn't a 5-year horizon anymore. It's happening now.