Google created the Tensor Processing Unit, a chip designed for machine-learning workloads.
Google developed the TPU as an application-specific integrated circuit for accelerating neural-network calculations. The company disclosed the technology publicly in 2016, explaining that TPUs had already been used inside Google data centers for products such as Search, Street View, and voice recognition.
Unlike a general-purpose central processing unit, a TPU is optimized for the matrix operations common in machine learning. Google has used successive TPU generations for both internal services and cloud customers, making specialized computing hardware an important part of its technology strategy.
The question identifies Google rather than Alphabet because the TPU was developed by Google's engineering organization. Alphabet became Google's parent company in 2015, and Google remains its principal internet and cloud subsidiary. The rise of AI computing has made custom accelerators a major factor in how investors assess large technology companies.