Apple’s server comeback depends on enterprise support—and perhaps Nvidia’s networking

Apple is developing an enterprise AI server built around its planned M8 Ultra chips, according to The Information. One design clusters two chips and another four; the intended buyers are AI developers, companies and governments running inference on their own equipment. Apple has discussed using Nvidia’s NVLink Fusion to connect the processors because its existing interconnects become too slow or expensive at server scale.

The project began roughly a year ago with the support of John Ternus, then Apple’s hardware chief and now its CEO. A launch is tentatively aimed at 2029, which makes every specification provisional: Apple could change the design, proceed without Nvidia or cancel it.

There is real demand behind the idea. AI laboratories have been buying Mac minis and Mac Studios in large quantities, while Apple’s Mac revenue rose 29% to $10.4bn in its latest reported quarter. A purpose-built server could convert that improvised use into a supported enterprise product and extend Apple silicon beyond consumer devices.

But Apple left this market once before. Xserve, sold from 2002 to 2011, suffered not only from limited scale but from weak enterprise software, remote management and customer support. Those gaps remain relevant. Apple’s MLX framework has a small team, and its former leader Awni Hannun has left for Anthropic. Nvidia’s role is also strategically revealing: even if alternative chips gain ground, Nvidia can collect revenue from the switches, chiplets and software that make accelerators operate as a system. Networking already represents a material share of AI data-centre costs and Nvidia’s data-centre revenue.