Apple, Foxconn, and Lenovo have been entrusted with the construction of M4-based AI servers.

November 9, 2024

It has been revealed that Apple is partnering with Foxconn and Lenovo to advance plans for building a new AI server based on Apple Silicon. According to Nikkei Asia, Apple has requested both companies to manufacture AI servers to be used for Apple Intelligence services in their own datacenters, in Taiwan. This initiative is believed to help Apple reduce dependence on China and enhance competitiveness in the field of AI.

It seems that Apple hopes to utilize Foxconn's Taiwan facilities to produce AI servers equipped with the Apple Silicon M4 chip.Foxconn is globally known for manufacturing AI servers based on NVIDIA GPUs (including H100 and H200), and their engineering capabilities and research and development resources are believed to significantly contribute to the construction of AI servers.The background behind Apple seeking production in Taiwan includes leveraging Foxconn's specialized technology and intending to supplement Apple's lack of in-house server design experience.

Foxconn has AI development bases in Hsinchu, Taiwan, and San Jose, USA, and is also working with NVIDIA to develop the next-generation AI server (GB300). However, the current production capacity in Taiwan is limited, necessitating preparation for future expansion. Therefore, Apple is considering not only collaborating with Foxconn but also with LCFC under Lenovo to complement server design. In anticipation of the increasing demand for servers accompanying the advancement of AI technology, Apple is exploring the establishment of a multinational production network including Foxconn, LCFC, and others.

Apple's AI server development differs from the approaches of competitors such as Amazon, Google, and Microsoft. While these companies develop high-performance GPU servers primarily for training large-scale machine learning models, Apple's strategy focuses mainly on providing AI functions specialized for user inference. As a result, Apple's servers do not require advanced cooling technologies like liquid cooling, enabling a simpler design.Apple's in-development Apple Intelligence is a suite of AI services for generation-oriented services on iPhone, iPad, and macOS, with an expected increasing need for servers as it expands.

Furthermore, Apple's server demand is limited and does not require mass production on the scale typically handled by Foxconn and Lenovo for the NVIDIA GB200 server. However, with the development of Apple's in-house AI, there is a possibility of expanding production scale, making Apple an attractive customer for Foxconn and Lenovo as well.

Apple's collaboration with Foxconn and Lenovo to build AI servers represents a solid approach towards internal use of next-generation AI technology. By leveraging the power of Apple Silicon for AI servers and specializing in inference processing, it is believed that efforts are made to optimize energy efficiency and costs. It is also noteworthy that by producing in Taiwan, Apple aims to reduce dependence on China and promote diversification of the supply chain. In the increasingly competitive AI field, how effectively Apple can expand AI inference infrastructure to enhance user experience will directly impact the company's competitiveness.