Definition
A graphics processing unit (GPU) is a processor built for massively parallel computation. It was created to break each on-screen frame into thousands of pixels and vertices computed simultaneously, and now also carries general-purpose workloads such as AI training, inference, and scientific computing.
How it works
A GPU combines thousands of lightweight cores that execute large batches of similar arithmetic tasks in parallel, trading latency for throughput. Paired with high-bandwidth memory, it excels at matrix math — the shared core of graphics transforms and neural networks.
Usage
Gaming, 3D rendering, and video encoding are the GPU’s traditional territory; deep learning training and inference have made it the center of AI infrastructure. Everyday users touch a GPU indirectly through video decoding in browsers and on-device machine learning features.
Related and confusable terms
Compared with a CPU, a GPU has many simple cores suited to parallel work rather than branch-heavy logic; a TPU is a tensor-focused ASIC with a narrower purpose. In the cloud, a virtual machine can be assigned a vGPU so that several instances share one physical GPU.