Tianfeng Securities analyst Mingzhi Guo disclosed on Tuesday that Apple Inc. will officially mass-produce its self-developed server AI chips in the second half of 2026 and plans to launch new data centers equipped with these chips in 2027. This timeline indicates that Apple expects a significant increase in AI demand on devices by 2027.

Apple's investment in the AI infrastructure field is accelerating. In February 2025, the company announced a four-year $50 billion U.S. manufacturing plan, which includes building a factory in Houston, Texas, dedicated to producing AI servers. The factory started operating ahead of schedule in October last year and has begun shipping servers manufactured in the U.S. for the Apple Intelligence platform.

GPU chip (3)

In terms of chip development, Apple launched an internal project codenamed "ACDC" in May 2024, specifically for developing Apple Silicon for AI server cluster processing. In December of the same year, Apple partnered with Broadcom to develop a processor called Baltra, which is planned for release in 2026. Broadcom's involvement may focus on "chiplet" technology, which involves combining multiple small chips into a larger one.

Currently, Apple's private cloud computing already uses servers with self-developed chips. Software chief Craig Federighi once stated that this design enables the private cloud to process queries in end-to-end encrypted mode. After transitioning to dedicated AI server chips, Apple is expected to gain multiple advantages, including improved performance, energy efficiency, and better thermal management.

In addition to the Houston factory, Apple will also expand data center capacity in North Carolina, Iowa, Oregon, Arizona, and Nevada. The new data center mentioned by Mingzhi Guo in 2027 seems to be specifically for AI processing, serving as an additional layout built upon existing data centers.