A custom-built AI chip from Google that was introduced in 2016 and used in Google data centers. Unlike the NVIDIA GPU, which was originally designed for graphics processing, the Tensor Processing Unit (TPU) was designed entirely for matrix multiplication in both AI training and inference stages. Running software from Google's TensorFlow AI library, TPU circuits process low-precision math very quickly (see
TF32).
In 2025, Google introduced Ironwood, which was enhanced for inference, although still used for training. Liquid cooled, up to 9,216 Ironwood TPUs can be linked with Google's InterChip Interconnect (ICI).
Major Changes
In 2026, Google split its 8th-generation TPU into separate training and inference chips (TPU8t and TPU8i). Also in 2026, a decade after the TPU's introduction, Google began selling its TPUs to other AI companies, and two of its first customers were Anthropic and Meta. The cost of TPUs is much less than NVIDIA's GPUs, causing NVIDIA to make changes such as ramping up product cycles. See
Google Cloud Platform and
Tensor core.
TPUs in a Smartphone
Starting in 2019, Pixel phones began using Edge TPUs, which were later integrated into the phone's primary chip (SoC). See
Tensor chip.