Nobody really knows what demand for AI infrastructure will look like.
The market is pricing today’s use cases, not what has yet to be invented.
Analysts are modelling compute demand around agentic AI and LLMs, forgetting that agents were barely part of the conversation three years ago, and that LLMs are just a tiny fraction of what Artificial Intelligence / Machine Learning is. As if agentic workflows are the final frontier.
One breakthrough could multiply demand for memory, compute, storage, and energy.
The best bet is probably owning the dominant companies that benefit no matter what gets built.
Author profile
Co founder Plaza Signup Here 👉 readplaza.com/i/chris
12 posts. 16 followers. 24 following.
Primary publisher: Hardwired
Public comments
Marc Atasoy · Jul 16, 2026
The mistake is assuming tomorrow's infrastructure demand will look like today's workloads. We tend to extrapolate instead of anticipate. But absolutely, the market dominators today will certainly be key players for whatever is rehsaping tomorrow.
