09 · Hardware & Infrastructure
GPU
Graphics Processing Unit
Chips first developed for graphics that also handle many neural-network calculations in parallel. Nvidia's hardware and CUDA ecosystem are widely used for AI training, while AMD and purpose-built accelerators also compete.
Concrete example
A frontier model is trained across tens of thousands of these chips wired together.
Why it matters
The pick-and-shovel of the AI gold rush.
Nearby in Hardware & Infrastructure
Evidence
Sources for this definition
- 1CUDA platform
NVIDIA · Documentation · checked 2026-07-13
- 2ROCm software
AMD · Documentation · checked 2026-07-13