AI is moving from a chip-led buildout into full-stack AI factories, shifting value capture toward system bottlenecks such as networking, interconnect, optics, memory, packaging, power, cooling, rack deployment, and enterprise data infrastructure.

As AI infrastructure shifts from training-dominated compute bottlenecks toward inference-heavy cluster scaling, photonics and interconnect become a standalone investable stack because the binding constraint moves to data movement, optical bandwidth, and bandwidth-per-watt.

Physical AI and humanoid robotics could create a future chokepoint stack across materials, actuators, specialty alloys, magnets, sensors, and power components, but the investable edge today is in identifying which inputs become binding at scale rather than assuming the whole robotics chain rerates immediately.