Apple is developing an enterprise server built around its own high-performance silicon, according to a report that has rippled across the technology and financial press this week — a move that would return the company to the server market it abandoned more than a decade ago and that would lean heavily on the same M-series Ultra chips now powering Mac desktops.
The product, first detailed by The Information and subsequently picked up by Ars Technica, Yahoo Finance, The Next Web, Seeking Alpha and TechPowerUp, is said to be planned for a 2029 release in two configurations: one with two of Apple's future M8 Ultra chips and one with four.
What Apple is reportedly building
According to the reporting, the machine would be a rack-scale enterprise server rather than a consumer appliance, targeted at companies running AI training and inference workloads. The project reportedly began about a year ago and had the backing of John Ternus, who at the time led Apple's hardware engineering organization and who, per the report, has since taken over as chief executive.
The four-chip configuration would represent a significant escalation of Apple's unified-memory architecture, which lets CPU and GPU cores draw on the same high-bandwidth memory pool without copying data across a PCIe bus. That design is the principal reason Apple's desktop chips have found an unexpected second life in AI work: large models can be held in memory that, on comparable discrete-GPU setups, would require several cards and far more power.
Why now: developers already voted
Apple's interest did not emerge in a vacuum. Sales of the Mac mini and Mac Studio have surged as AI developers and smaller labs have snapped them up to run local models, fine-tuning jobs and inference pipelines. The appeal is straightforward — comparatively low cost per gigabyte of usable unified memory, quiet operation, and a power envelope that fits under a desk.
That grassroots adoption gives Apple something it rarely has in the data center: a proven demand signal. An enterprise server using the same chips would let the company sell into organizations that have already validated Apple silicon for AI work but want denser, rack-mountable hardware and vendor support.
The Nvidia wrinkle
Perhaps the most surprising element of the report is that Apple is said to be turning to Nvidia — not for GPUs, but for networking technology. Several outlets framed the story specifically around Apple tapping Nvidia's networking equipment to connect the servers, a detail that matters because the two companies have a long and frosty history dating back to a 2008 chipset dispute, after which Apple moved its graphics business to AMD.
The arrangement, if accurate, reflects the reality of modern AI infrastructure: compute is only half the problem, and Nvidia's interconnect and networking portfolio — built out through its Mellanox acquisition — is effectively the industry standard for stitching accelerators together.
A return after nearly two decades
Apple's last serious server push was the Xserve, a 1U rack machine introduced in 2002 and discontinued in 2011, with the Xserve RAID storage line killed off in 2008. The company has since repeatedly insisted it has no interest in the enterprise data-center business.
A 2029 launch would put roughly 18 years between the Xserve's demise and Apple's return — which is why the framing across outlets has been consistent on one point, captured by Ars Technica's description of the project:
"The potential product's expected release in 2029 would mark the first Apple server to hit the market in nearly two decades — and could capitalize on the surging popularity of Apple hardware among AI developers."
It is worth distinguishing this reported product from Apple's existing server-class work. The company already runs Apple Intelligence requests on its own silicon in data centers through its Private Cloud Compute system, announced in 2024. That infrastructure is internal and privacy-focused; the M8 Ultra machine described in the new report would, by contrast, be a commercial product sold to enterprises.
How the story was framed
- Ars Technica led with the hardware story and the developer-adoption angle, emphasizing the Mac mini and Mac Studio sales boom as the commercial justification.
- Yahoo Finance cast it as an "enterprise comeback push," foregrounding the Nvidia partnership and the market implications.
- TechPowerUp headlined the Nvidia networking angle, treating Apple's return to servers as the secondary hook.
- The Next Web kept the framing narrow and factual: Apple is reportedly developing an enterprise server built on its own chips.
- Seeking Alpha approached it from the investor side, flagging the in-house silicon plus Nvidia networking combination as a potential new revenue line.
Skepticism and open questions
Not everything in the report is easily reconciled. A 2029 ship date is unusually far out for a product to surface in supply-chain reporting, and Apple has a well-documented habit of killing internal projects — including, reportedly, its own high-end Mac Pro follow-ups and the long-rumored Apple Car. Chip roadmaps that far ahead are also speculative; an "M8 Ultra" is several generations beyond anything Apple has announced.
There is also the question of software. Apple's ecosystem is built around Metal rather than CUDA, and while PyTorch and other frameworks have added Metal support, most enterprise AI stacks are tuned for Nvidia's software layer. An Apple server would need to overcome that gravity, or find buyers whose workloads are already Metal-native.
Still, the strategic logic is hard to dismiss. Apple has spent a decade building chip design capability that now rivals anyone's, has a captive developer base that has proven willing to use it for AI, and faces a data-center market desperate for alternatives to a single dominant supplier. If the reporting is accurate, the company appears to have concluded that the fastest route back into the server room runs through the silicon it already sells — and, for the plumbing, through the competitor it once walked away from.
