Apple has reportedly started developing custom Apple silicon chips for AI processing on the company’s data center servers. The effort, codenamed Project ACDC, aims to deploy these AI chips to power future AI features that run online, according to a new report from The Wall Street Journal.

Apple is collaborating closely with TSMC on the project, seeking to leverage the Taiwanese chipmaker’s advanced nodes to create efficient AI chips. The project has reportedly been in the works for several years, though the report cautions that a release date for the AI chips is yet to be locked down, indicating that an unveiling is unlikely to happen anytime soon.

Apple plans to utilize the AI chips to run server-side AI models rather than to train them. Training AI models will likely remain reliant on NVIDIA’s chips in the foreseeable future. With Project ACDC, Apple hopes to leverage its silicon expertise to advance its AI ambitions, aiming to gain an edge over other players in the AI market like Google, Meta, Microsoft, and OpenAI. While Apple ultimately plans to run most of its future AI capabilities on-device, some comprehensive AI features still require some form of cloud processing, which is where Apple’s planned chips would step in.

Apple’s cloud AI chips would represent an important step in the vertical integration of the company’s overall AI efforts, with these chips supplementing the Neural Engine already present on most Apple products. The report also adds that high-end variants of unreleased Apple silicon, likely the M4, could power Apple’s AI data centers in the future. However, dedicated AI chips seem more likely if Apple goes forward with the project.

Apple is rumored to unveil the M4 chip later today with the next-generation iPad Pro at its “Let Loose” event. While the company is unlikely to discuss software AI features today as it keeps them for WWDC 2024, Apple could discuss the AI hardware capabilities of the new processor.