Is AI’s meteoric rise about to be tempered by some practicalities? Many have predicted that AI’s energy consumption is unsustainable, but that may be just the tip of the iceberg.

The computing industry loves a trend. Even better if it’s one that’s exponential. But those trends remain true for only so long before they suddenly stop dead or even go into reverse. As long as two decades ago, Intel co-founder Gordon Moore sounded the alarm on the trends that underpinned his eponymous law. “No exponential is forever,” he reminded engineers at a chip industry conference in 2003. Within a year, one major driver of silicon scaling – improvements in power efficiency as transistors grew smaller – had run into the sand. And microprocessor clock speeds have remained pegged at a few gigahertz ever since.

Today, Moore’s law looks even wobblier. The cost reductions promised every two years to support a doubling in transistor and memory density are no longer realistic. For the revival of artificial intelligence...