Oracle Cloud Infrastructure to host Google TPU accelerators

Oracle and Google have agreed to expand their cloud partnership by bringing Google's custom artificial intelligence chips to Oracle Cloud Infrastructure. Starting in 2027, Oracle will become the first major third party cloud provider to host Google's Tensor Processing Units, commonly known as TPUs.

For years, tech companies building artificial intelligence have relied heavily on standard hardware from Nvidia. Google originally built its own custom processors to cut costs and boost performance for massive computer learning tasks. Under the deal, Oracle will supply the data center infrastructure, power, and cooling needed to run Google's newest eighth generation chips for its customers.

This move gives software developers more freedom to choose the right silicon for their specific computing tasks without being locked into a single provider. Greater competition among hardware suppliers typically reduces computing costs, which can ultimately lower prices for consumer facing AI applications.

However, a multi year wait brings uncertainty. It is still unclear whether chip availability by 2027 will arrive quickly enough to capture current demand, or if competing hardware releases will shift customer preferences in the interim.