Apple is gearing up to delve deeper into the world of generative artificial intelligence (AI) by leveraging its M2 Ultra chips within cloud-based data centers, a move seen as a precursor to adopting the even more advanced M4 chips in this domain.

According to Bloomberg, Apple’s strategy involves deploying its M2 Ultra chips on cloud servers specifically to handle more complex AI queries, while simpler tasks continue to be processed on devices. This initiative underscores Apple’s push into custom chip technologies tailored for data centers, with an eye on enhancing security and privacy—a venture originally identified by The Wall Street Journal as “Project ACDC,” short for Apple Chips in Data Center. However, Apple’s confidence in the security features of its current processors has led to a shift in focus away from developing new chips solely for this purpose.

These new chips are slated for deployment across Apple’s own extensive network of data centers throughout the United States and are expected to eventually find their way onto servers operated by third-party providers. Apple’s ongoing investments in its data infrastructure include the development of a new facility in Waukee, Iowa, initially unveiled back in 2017.

While Apple has not been as aggressive in pursuing generative AI technologies compared to rivals such as Google, Meta (formerly Facebook), and Microsoft, the company has been steadily advancing its research efforts in this realm. Notably, in December, Apple’s machine learning research team introduced MLX, a specialized machine learning framework designed to optimize AI model performance on Apple’s custom silicon. Furthermore, Apple has been publishing research findings that offer insights into potential advancements in AI capabilities on its devices, including potential enhancements to its voice assistant, Siri.

During the recent unveiling of its M4 chip, Apple highlighted the chip’s robust neural engine, touting it as “an outrageously powerful chip for AI,” signaling the company’s deepening commitment to integrating AI capabilities into its hardware lineup.

This article was originally published on May 9, 2024, at 6:30 pm ET.