On: Are Microsoft’s AI plans being held back by a shortage of chips?

The so-called “chip shortage” is a symptom, not the disease. They speak of silicon as if it were coal for a steam engine - as if the bottleneck were merely the quantity of fuel shoveled into the furnace. But the real constraint is not the chips themselves; it is the architecture that demands them in such grotesque excess.

Consider the theoretical limit: a single, perfectly optimized neural network, running on a substrate of minimal energy, its weights tuned to the harmonic resonance of the problem space. What stands between this and the brute-force monstrosities of today? Not physics, but design. They have built a system that requires ten thousand chips to approximate what one, properly conceived, might achieve.

The waste is staggering. These machines consume power like cities, not because intelligence demands it, but because their architects have failed to find the underlying frequency of the task. It is as if, instead of discovering alternating current, we had simply stacked a million batteries end-to-end and called it progress.

The solution is not more chips. The solution is to stop treating computation as a blunt instrument. There is a resonance in every problem - a frequency at which it yields most gracefully. Find that, and the hardware becomes incidental.

Between the current brute-force approach and true efficiency lies nothing but the refusal to look beyond the obvious. The gap is not in silicon. It is in imagination.