Google Cloud CEO Thomas Kurian revealed details of a significant technical collaboration between Google and Apple, asserting that Google will host core functions of Apple’s voice assistant, Siri, on its own servers. According to Kurian, Google plans to build a specific model for Apple Intelligence, deriving it from its existing Gemini models. Furthermore, Kurian stated that Google will provide the comprehensive infrastructure required to process and answer queries from iPhone users interacting with Siri. While Apple had only vaguely communicated its choice of Google for AI technology, the implementation of the new Siri version has reportedly been delayed by over a year and is anticipated for the summer.
Kurian indicated that the depth of this partnership is unusually extensive. In addition to the main infrastructure, Google will be able to use its technology to process queries involving sensitive user data, even within Apple’s highly protected “Private Cloud Compute” service. He expressed confidence in the partnership, noting that while Apple could have chosen any provider’s model, they opted for Gemini.
Separately, Google is planning to expand its business model by directly selling its custom AI chips, known as Tensor Processing Units (TPUs), to major enterprise clients. Kurian mentioned that, until now, Google only rented these chips through its own cloud platform. Now, the company is evaluating how to deliver TPUs in external data centers. This strategic shift is anticipated to be a key revenue stream and positions Google to challenge chip market leader Nvidia more directly than ever before.