Apple is reportedly considering a return to the server market it left when it abandoned Xserve, but this time incorporating the future M8 Ultra chip, fully integrated with NVLink Fusion technology.
While Apple already makes its own servers in Houston, it is reportedly looking at making a rack-mounted server product for wider distribution.
The rumored product line is said to include two main configurations. One would feature two M8 Ultra chips, while the other would use four of the still-unreleased processors.
As part of the technology assessment, Apple is said to be considering using NVLink Fusion. That is high-bandwidth, low-latency connective technology that allows companies to plug chips and AI accelerators directly into Nvidia's advanced data center infrastructure.
Originally limited to just Nvidia hardware, the company has recently opened up the technology for others to license and use.
Nothing will hit the market before 2029, and it's not clear exactly who can buy what will be ridiculously expensive hardware. A report by The Information on Wednesday morning discusses that plans are to sell to "AI developers, other businesses and governments," with the latter two the main target of the original xServe hardware.
New CEO John Ternus is said to have been a backer of the project, when it sprung up inside Apple about a year ago. And it would need his full backing, as the effort would have to source expensive components, in epic quantities, well before the project starts.
Apple refuses access requests to its self-manufactured Private Cloud Compute hardware. The company has also partnered with Google and Amazon for cloud capacity in the past.
Apple and Nvidia
The report paints Apple's relationship with Nvidia as "frosty," but I disagree with that sentiment. It's been beyond that, and publicly, it's been downright frozen.
That relationship started with Steve Jobs in 2001. It led to defective solders, poor performance, and allegations of trade secret theft made by Jobs.
Apple support for Nvidia hardware has been poor for over a decade. There were drivers for Nvidia cards in Intel hardware, but that effectively stopped after the 5,1 Mac Pro stopped shipping, and before Apple popularized eGPU enclosures in 2016.
By 2019, video card driver support was killed, for good.
But, perhaps, there's a thawing now. Money talks, after all.
"We would always rather sell something than nothing," Dion Harris, Nvidia's senior director of high-performance computing and AI infrastructure, said in an interview. Reportedly, Harris declined to comment on whether Apple is an NVLink Fusion customer, but said that "Apple has always been a great customer and partner of ours."
That partnership and customer status may be recent. Apple was said in March 2026 to have bought $1 billion on Nvidia hardware for research. Specifically, it was said to have purchased 250 Nvidia NVL72 servers at a cost of between $3.7 million and $4 million each.
For a look back at the Xserve, Jeff Geerling has an excellent saga about refurbishing the 20-year-old G5 version of the hardware. Notably, this hardware was used in the Virginia Tech "Big Mac" cluster in the summer of 2004, after launching in 2003 with towers.
And I saw them again about four years later at the university's surplus auction, where they turn up from time to time to this day.
