Watts and wafers may define the AI trade

Gavin Baker’s AI thesis comes down to two words:
Watts and wafers.
That may be the cleanest way to understand why the AI trade is different from a normal software cycle.
The bottleneck is not just code.
It is power, chips, fabs, memory, data centers, and the physical infrastructure needed to train and run frontier models.
That is why names tied to compute supply, energy availability, and semiconductor capacity keep mattering.
For $NVDA, $TSM, $MU, and the broader AI infrastructure stack, the question is no longer whether demand exists.
The question is whether the world can build enough physical capacity fast enough.