A new report suggests Apple is developing an AI inference server that could use multiple M8 Ultra chips and possibly Nvidia's NVLink Fusion technology, targeting a 2029 launch aimed at AI developers, businesses and governments.
Apple may be preparing one of its most significant hardware shifts in years: a comeback in the enterprise server market, this time built around artificial intelligence. According to a report by The Information, the company is working on an AI-focused server that could run on two or four future M8 Ultra chips, aimed at AI developers, businesses and governments, with a possible launch in 2029. Apple has not confirmed the project, and it could still be altered or scrapped before reaching the market.
The reported effort comes as AI companies increasingly turn to Apple's high-performance Mac hardware to run AI workloads, a trend that appears to have shaped the server's design goals.
Inference-focused design, Nvidia in the mix
The server would reportedly come in two configurations, one pairing two M8 Ultra chips and a larger version combining four. Its purpose would centre on AI inference, meaning running already-trained models and generating responses, rather than training large models from scratch. The project is said to have begun roughly a year ago, when John Ternus led Apple's hardware engineering division; Ternus became Apple's CEO on September 1.
The timing lines up with reported demand from AI firms for Apple's smaller machines. The Information reported that OpenAI had purchased tens of thousands of Mac mini and Mac Studio units, while Anthropic has rented Mac minis through Amazon Web Services.
Perhaps the most unexpected element of the report is that Apple is reportedly considering Nvidia's NVLink Fusion to link its M8 Ultra chips together. NVLink Fusion combines hardware and software built to let processors communicate at very high speeds within data centres, and Apple is said to believe it could provide the connectivity needed to scale multiple chips inside a single AI server. The prospect stands out given the long-strained relationship between Apple and Nvidia, and any such talks would point to AI infrastructure opening new ground for cooperation between the two. Nvidia's involvement, however, remains unconfirmed, and Apple could move forward without it.
A challenging comeback to enterprise hardware
Should it launch, the machine would mark Apple's return to dedicated enterprise servers, a market it exited when it discontinued the Xserve line in 2011. Apple already runs servers for its own Private Cloud Compute system, but those are not sold externally. A commercial AI server would be different, sold directly to outside organisations.
That shift brings its own hurdles. Enterprise buyers typically need specialised software, remote management tools, developer support and long-term servicing, meaning Apple would need to build far more than hardware alone. For now, the M8 Ultra server remains an unconfirmed, reported project. If it does arrive by 2029, Apple would be entering an AI server market currently dominated by Nvidia, potentially using both its own silicon and Nvidia's own technology to compete.
