Artificial Intelligence, machine learning are intrinsically linked to Google. So much so that it custom built a TPU for this very purpose.

Although earlier Google has built hardware for its data centres, but to actually go ahead and design a customized processor for its own is perhaps an indicator as to the extent the search engine giant is willing to go to get the results its seeks.

Towards this end, Google has now revealed that it has designed the Tensor Process Unit, a custom built chip specifically designed for machine learning. Since Artificial intelligence operations typically make use of parallel processes, the TPU has been designed keeping that in view.

According to Google, it’s an “order of magnitude” faster in Artificial Intelligence related operations than conventional processor consuming the same amount of power. Not only is it power-efficient, its form factor optimizes on space too: it can fit into the hard drive bays of a rack in a typical data centre.

As for the processor’s capability? Well, as you may be well aware of, Google has been making extensive use of TPUs lately. It has used them from improving the quality of its maps to securing AlphaGo’s victory over its human champion.

The fact that the TPU enabled AI could predict much faster and further than its human rival essentially led it to its victory. Although Google is not selling the TPU, you can nonetheless benefit from it thanks to Google’s ever expanding services.