Apple is working on an enterprise AI server built around its own chips, which would be its first server for outside customers since it retired the Xserve in 2011. The machine would come in two configurations, with either two or four of Apple’s future M8 Ultra chips, and could be released in 2029, The Information reported, as summarized by Ars Technica.
The project began about a year ago with backing from John Ternus, who was leading Apple’s hardware engineering at the time and took over as CEO from Tim Cook on Sept. 1. It could still be canceled.
An Nvidia link
To tie several M8 chips together at data-center scale, Apple is considering Nvidia’s NVLink Fusion interconnect, according to the report, and the two companies have discussed how Apple could use Nvidia technology in its AI push. The Information cautioned that the server could go ahead without Nvidia’s technology. Reuters also picked up the Nvidia angle.
The timing follows a run on Apple desktops for AI work. AI developers and companies have been buying Mac minis and Mac Studios to run models, and The Information reported that OpenAI has bought “tens of thousands” of them to train agents through reinforcement learning, while Anthropic has rented Mac minis from Amazon Web Services. Apple’s M-series Ultra chips pair high performance with a large pool of unified memory, which suits running big models locally.
Apple already builds servers for its own Private Cloud Compute, the infrastructure behind cloud-based Apple Intelligence features. Earlier this year, 9to5Mac reported that some of those servers were sitting unused in warehouses because Apple Intelligence usage was lower than expected. Selling servers to enterprises would give that silicon a much larger market.
Supply could be the hard part. The industry-wide memory shortage driven by AI data-center construction has already pushed up prices on consumer electronics, including Apple products, Ars noted.
WebProNews is an iEntry Publication