What is a GPU, and why does AI need it?
Short answer
A graphics processing unit (GPU) is a chip with thousands of small cores that perform many calculations in parallel. It was built for graphics but is now the main hardware for training and running AI models.
CPU versus GPU
A central processing unit (CPU) has a few powerful cores designed to run varied tasks one after another very quickly. A GPU has far more, simpler cores designed to run the same operation on large blocks of data at once, which suits drawing the pixels of an image.
Why AI uses them
Neural networks are largely made of matrix multiplications, which can be split into many independent calculations. That structure matches a GPU’s parallel design, so training and running AI models on GPUs is much faster than on CPUs. Memory capacity, memory speed and the links between chips matter as much as raw compute; high-bandwidth memory (HBM) is a key component of AI accelerators.
Who makes them and alternatives
Nvidia, AMD and Intel design GPUs. Companies including Google (Tensor Processing Units) and Amazon design custom AI chips, which are a type of ASIC (application-specific integrated circuit). Designers mostly do not manufacture their own chips; foundries do, which is why advanced manufacturing capacity features in many AI and chip stories.