GPU
A chip built to do enormous numbers of simple calculations at once. Training and running large models depends on them.
What it tells you about a developer
A candidate who talks about GPU cost, memory, and utilization has run models rather than only called them. That work sits close to infrastructure, and it is a smaller pool than product AI engineering.
GPU stands for graphics processing unit. The chips were built for video game graphics, and the same design turned out to suit the math behind machine learning. Nvidia makes most of the ones AI companies buy. GPUs are scarce and expensive, which is why access to them shapes what a company can train, and why cloud providers rent them by the hour.