Behind the Google products you use every day are custom chips designed for one task: doing math at scale. They are called TPUs, or Tensor Processing Units.
We've built TPUs from the ground up over the past decade to use AI models. Basically, it takes a lot of calculations for AI models to work, and TPUs can perform complex calculations very quickly: The new generation of TPUs can process 121 exaflops of computing power with twice the bandwidth of previous generations.
Learn more about these small but powerful processors in the video below.
