GPU
Graphics Processing Unit
ELI5 — The Vibe Check
A GPU was originally built for rendering graphics in games, but turns out it's also perfect for AI. GPUs can do thousands of simple math operations simultaneously — exactly what neural networks need. Training a big model on a CPU takes months; on a GPU, it takes days. NVIDIA's GPUs are basically the oil of the AI economy — everyone needs them and there's never enough.
Real Talk
GPUs are parallel processing units optimized for matrix operations, making them ideal for deep learning training and inference. NVIDIA dominates the AI GPU market with data center GPUs (A100, H100, B200) featuring high-bandwidth memory and tensor cores. AMD (MI300X) and Google (TPUs) provide alternatives. GPU access is the primary bottleneck in AI development.
When You'll Hear This
"We need at least 8 H100 GPUs to train this model." / "GPU prices are insane right now — everyone wants them for AI."
Related Terms
Cloud Computing
Cloud computing means using computers that live in someone else's giant warehouse instead of your own machine.
CUDA
CUDA is NVIDIA's secret weapon — it's the programming platform that lets developers use NVIDIA GPUs for AI, not just gaming.
Inference
Inference is when the AI actually runs and generates output — as opposed to training, which is when it's learning.
Neural Network
A neural network is a system loosely inspired by the human brain — lots of little math nodes connected together, passing numbers to each other.
Training
Training is the long, expensive process where an AI learns from data.