Apple plans to start its foray into the generative AI field by offloading complex queries before its advanced M4 chip onto the M2 Ultra chip running in data centers.

According to Bloomberg, Apple plans to deploy its M2 Ultra chip on cloud servers to handle more complex AI queries, while simple tasks will be processed on devices. Previously, The Wall Street Journal reported that Apple wanted to manufacture custom chips and bring them into data centers to ensure security and privacy, a project referred to as Project ACDC. However, the company now believes that its existing processors already have sufficient security and privacy components.

These chips will be deployed in Apple's data centers and eventually on third-party operated servers. Apple operates its own servers across the United States and announced the construction of a new data center in Waukee, Iowa in 2017.

Although Apple has not progressed as quickly as competitors such as Google, Meta, and Microsoft in the generative AI field, the company has been conducting research in this technology area. In December of last year, Apple's machine learning research team released MLX, a machine learning framework that enables efficient operation of AI models on Apple silicon. The company has also published other research on AI models, which suggests what AI on Apple devices might look like and how existing products like Siri could be upgraded.

When announcing the new M4 chip, Apple emphasized the importance of AI performance, calling its new neural engine "an astonishingly powerful AI chip."