We evolved for a linear world, but AI development is exploding exponentially. From 2010 to now, the amount of training data for frontier AI models has grown by a staggering 1 trillion times. This growth is driven by advances in computational power, memory, and connectivity.

Despite skeptics predicting walls, the combined forces driving this revolution make the exponential trend predictable. To understand why, consider AI training as a room full of people working calculators. Today’s revolution is about ensuring those calculators never stop and work together as one.

Three advances are converging to enable this: faster calculators, faster data transfer, and the ability to connect hundreds of thousands of GPUs into warehouse-size supercomputers. These gains deliver dramatically more compute, with training times decreasing from 167 minutes to under four minutes on equivalent modern hardware.

This growth is not limited by Moore’s Law, which would predict only a 5x improvement. Instead, we’ve seen a 50x improvement, driven by the convergence of these advances. As AI development continues to accelerate, it’s clear that the exponential growth won’t hit a wall anytime soon.

Photo by Hannah Nelson on Pexels
Photos provided by Pexels