It is easy to talk about AI as if the model is the whole product. I find the broader stack more interesting.
There is hardware underneath the compute, infrastructure underneath the models, software around the models, and then applications that turn those capabilities into something useful for people. Improvements at one layer create opportunities and constraints at the layers above and below it. As AI scales, understanding those interactions becomes increasingly important.
I think this is particularly relevant for people building products. The application is where users experience AI, but what is possible at that layer depends heavily on what is happening underneath it.
Better models can change the interface. Cheaper inference can change the economics. More capable infrastructure can enable entirely different workflows. And sometimes a change at the hardware or systems level can eventually change what a product can do.
I don't think a PM needs to become an expert in every layer. But I do think there is value in understanding how the stack fits together, especially as the boundary between software, models, infrastructure, and the product itself becomes less clear.