The Gravity of Data in an AI World

In the early days of storage, we focused on "capacity per dollar." Today, we focus on "throughput per GPU."

AI has changed the gravity of data. You can no longer afford to have your dataset sitting in a cold tier while your compute clusters are spinning their fans. The bottleneck has shifted from the CPU to the storage fabric.

I’ve spent years looking at latency numbers, but AI is different. It’s a ravenous beast that requires data to be local, fast, and persistent. We are moving toward a world where storage is no longer a passive container, but an active participant in the training loop.