Jevons’ Paradox is real, but it is not evenly distributed.

Let's look at history.

When stream engine efficiency improved, it diffused.

Note: the actual technological innovation was steam engine, not coal. Yet Jevon's Paradox is about coal, not steam engine.

Early, static steam engine makers like Boulton & Watt didn’t become the kings of railroads and ships; instead, specialized locomotive and marine engines took over.

The same happened in computing: IBM ruled mainframes, but when compute demand exploded, the winners were Intel, Microsoft, and later, cloud hyperscalers and now nVidia. Yes, the actual technological innovation is around Moore's Law, not a particular manufacturer.

AI will get cheaper, smaller, and more ubiquitous. But is it going to be centralized in nuclear powered data centers? Jevon's Paradox won't tell you one way or another. One thing is clear, more AI won’t just mean more data center GPUs; it will mean AI everywhere, running on optimized silicon.

But unlike steam engines, computation is more general-purpose. If Nvidia adapts—by moving beyond data centers into edge AI, consumer hardware, and decentralized compute (their root) —it could remain dominant.

If it fails to pivot, AI’s diffusion could mirror history: demand soars, but the spoils go to new players better suited to the new paradigm.

Again, Jevons was right—growth is inevitable. But the second order intellectual question is: who captures it.