Apple weighs Nvidia tech for a possible return to AI servers

Apple is considering a return to selling enterprise servers, pairing its own chips with Nvidia networking technology, Reuters reported citing The Information. The project would target AI inference for outside customers, not only Apple’s internal Private Cloud Compute fleet.

Nothing is announced. The report says the effort could still be canceled or launched without Nvidia gear.

What the reported box looks like

According to coverage of The Information’s story, Apple is developing an AI server offered with either two or four planned M8 Ultra chips, its highest-performance processors. One option under discussion is Nvidia’s NVLink Fusion, a set of switches, chiplets, and software that help chips exchange data in data centers.

MacRumors notes Apple already builds servers for Private Cloud Compute, which handles some Apple Intelligence workloads off-device. Scaling Apple’s existing chip-to-chip links for a commercial product would raise cost and speed problems, the report says, which is why Nvidia’s stack is on the table. Apple has also declined partners who asked to buy those Private Cloud Compute machines.

Why 2029 and why Xserve nostalgia

The Information places a possible release no earlier than 2029. That would put Apple back in a market it left when it discontinued Xserve in 2011. The pitch, as retold by Reuters and MacRumors, is selling to AI developers, businesses, and governments that want to run models on their own hardware, with a focus on inference.

CEO John Ternus reportedly backed the project when it began about a year ago, while he was still Apple’s hardware engineering chief. Demand signals already exist in the wild: AI shops have been buying Mac mini and Mac Studio boxes in volume for local inference.

Partnership irony and open questions

An NVLink Fusion deal would put Nvidia inside an Apple product that could compete with Nvidia’s own AI systems, while extending Nvidia’s business of networking other companies’ silicon. Apple and Nvidia have a prickly history, but Apple has already said it will extend Private Cloud Compute onto Google Cloud using Nvidia GPUs, and the companies have explored other AI collaboration, MacRumors reported.

Software and go-to-market remain thinner in the reporting than the silicon. Coverage flags a need for stronger enterprise support and more AI developer tooling, including further MLX investment. Until Apple confirms M8 Ultra servers, treat the story as an early project leak with a long fuse, not a product roadmap.