Everyone talks about GPUs, power demand and data centers, but I think one of the more interesting long-term bottlenecks is being overlooked: **photonics**.
The basic idea is simple. As AI clusters get larger, the problem is not only compute. It is also data movement. GPUs, servers and racks need to communicate with each other at extremely high speeds, and copper starts to run into limits around distance, heat, signal loss and energy efficiency.
That is where light becomes interesting.
Photonics is already used in areas like optical transceivers, fiber networks, semiconductor manufacturing, sensors and lasers. But the investment case becomes more interesting when light moves closer to critical bottlenecks:
AI data centers need faster and more efficient interconnects.
Defence systems need better sensing, targeting and lower-cost response layers.
Satellite networks need high-bandwidth laser communication.
Advanced semiconductors already depend on light through lithography and inspection.
Packaging and testing may become the hidden bottleneck if optics move closer to chips.
I don’t think this is a simple “buy every photonics stock” theme. The sector is broad, definitions are messy, and many companies only have partial exposure. But the bigger point is that light may become an increasingly important infrastructure layer behind AI, defence, space and advanced chips.
Curious how others are thinking about this. Are photonics, optical interconnects and co-packaged optics still too early, or is this one of the more underappreciated infrastructure themes behind AI?
