AI Inference Leaves a Growing Storage Bill Behind

What happened

AI inference generates back-end data that can add significantly to storage costs, but those costs are often overlooked in infrastructure planning. WD’s internal measurements found that an 8-second AI-generated video returned as a 3.97MB 720p file left more than 500MB of associated back-end data, or 130x the output file size. McKinsey projects inference demand will grow at a 35% CAGR through 2030, faster than training demand.

Read next

• More changes in today’s briefingMicrosoft Research and Collaborator Resolve the KLS Conjecture With Human-Directed AI

The story so far

WD said AI inference generates back-end data that can substantially increase storage costs, citing an internal measurement in which an eight-second AI-generated video produced a 3.97MB 720p file but left more than 500MB of associated back-end data—about 130 times the output file’s size. WD said much of the data generated behind prompts, images and videos needs to be stored, creating costs that infrastructure planning often overlooks. The potential scale of that burden is tied to growing inference use: McKinsey projects inference demand will grow at a compound annual rate of 35% through 2030, faster than demand for training.