Google is developing its own artificial intelligence chips designed specifically to be used for machine learning, the company announced Thursday on the sidelines of the Google I/O conference.
Google revealed it has been using its own Tensor Processing Unit (TPU) chips for more than a year. The chips are specially created for the task of machine learning and require fewer transistors for operations, Google explains. It means they can accomplish more tasks faster.
The processors powered AlphaGo when it defeated the world's best human player of the Asian board game Go in March - "enabling it to 'think' much faster and look farther ahead between moves," according to Google.
Until that defeat Go had been considered too complex for a computer to play.
TPUs are also used for Google's Street View mapping and navigation service.
It was not previously known that Google was developing its own chips, and it could mean new competition for speciality firms in the semiconductor industry.
Nvidia is another company pursuing machine learning artificial intelligence technology. Apple has for some time designed its own processors for iPhones and iPads based on technology from chip developer ARM.