Infrastructure
Graphics Processing Unit
GPUThe specialized chip that does the heavy parallel math behind training and running AI models.
A GPU was originally designed to render graphics, but its ability to do thousands of calculations in parallel turns out to be ideal for the matrix math inside neural networks. GPUs are the workhorse hardware of modern AI, and access to enough of them is a real constraint on who can train large models. Their scarcity and cost are a major factor in the economics of the whole field.
Reference: Wikipedia: Graphics Processing Unit
Related terms
Tensor Processing Unit
A chip designed by Google specifically for machine learning, as an alternative to general-purpose GPUs.
Compute
The raw processing power used to train or run a model, a key driver of cost and capability.
Inference
Actually running a trained model to get an answer, as opposed to training it in the first place.