Home / Technology / Google's TPU: AI Chip Explained
Google's TPU: AI Chip Explained
29 Aug
Summary
- Google's TPU, or Tensor Processing Unit, is a proprietary AI accelerator.
- First deployed in 2015 for cloud data centers, it handles AI/ML calculations.
- Pixel 11's Tensor G6 chip uses TPU for camera, image, and local AI tasks.
Google first introduced its proprietary AI accelerator, the Tensor Processing Unit (TPU), internally in 2015. This specialized chip is designed to manage the intensive mathematical computations required for artificial intelligence and machine learning workloads, primarily within data center applications.
Recently, the term TPU has gained prominence due to its inclusion in Google's Pixel 11 smartphone, featuring the Tensor G6 chip. This mobile iteration of the TPU is described as handling camera and image processing, alongside local AI tasks, effectively acting as a Neural Processing Unit (NPU) for the device.
While the data center TPU and the smartphone TPU share similarities with other processors like NPUs and GPUs, they possess distinct characteristics. The dual application highlights the evolving role and naming conventions of AI-focused hardware in both cloud and personal computing environments.
