Apple is reportedly working on its own enterprise AI servers, built around future M8 Ultra chips, according to a September 16, 2026 report from The Information. If the project reaches the market, it would be Apple’s first dedicated server hardware since it discontinued the Xserve line in January 2011 — a gap of roughly 15 years.
- Report source: The Information, September 16, 2026
- Configurations: servers with either 2 or 4 M8 Ultra chips
- Use case: AI inference, not model training
- Possible interconnect: Nvidia’s NVLink Fusion (unconfirmed)
- Buyers: AI developers, businesses, governments
- Target launch: not before 2029; project not finalized
What the report says
The Information, citing people familiar with the plans, says Apple is considering two server configurations: one built around two M8 Ultra chips, and a larger version with four. Both are aimed at AI inference — running already-trained AI models — rather than training new ones from scratch.
To connect the chips inside each machine, Apple is reportedly in talks with Nvidia about using NVLink Fusion, an interconnect technology Nvidia originally built for its own GPUs and later opened up to outside chipmakers. No agreement has been announced, and the report notes Apple could still build the server without Nvidia’s technology — or shelve the project entirely.
Unlike Apple’s internal server fleet, this machine is reportedly meant to be sold. The Information says the target customers are AI developers, businesses, and governments running their own AI infrastructure — not Apple’s typical consumer base.
Why Apple needs more AI capacity
The timing lines up with a real capacity problem. Apple Intelligence and Siri reportedly run on repurposed M2 Ultra chips, and Apple has reportedly been renting additional Nvidia GPU capacity through Google Cloud to keep up with demand.
Apple already runs its own custom servers for a system called Private Cloud Compute, which handles AI requests too complex for an iPhone or Mac to process on-device. Those machines aren’t sold outside the company — the reported M8 Ultra server would be a commercial version of that same idea.
A leadership change adds context
The report comes two weeks into John Ternus’s tenure as Apple CEO. Ternus, who previously led Apple’s hardware engineering organization, took over from Tim Cook on September 1, 2026, when Cook moved into the newly created role of executive chairman — Apple’s first CEO change since 2011. A server project like this would fall squarely within the division Ternus used to run.
It also arrives the same month Apple launched the iPhone 18 Pro and its first foldable iPhone — a sign of how much is moving through Apple’s product pipeline right now.
What’s still unknown
A lot. Apple has not confirmed an M8 chip, a server product, or any Nvidia partnership. The Information’s report describes a project still being evaluated internally, with no pricing, official name, or release date beyond a rough “not before 2029” target. For scale, Apple’s current M5 Ultra chip supports up to 512GB of unified memory — but how memory would work across two or four linked M8 Ultra chips hasn’t been reported.
Nvidia hasn’t commented on the report, and NVLink Fusion has not been confirmed as part of any Apple product.
The bigger picture
Apple would be joining Amazon, Google, and Microsoft, which have all built custom AI chips rather than relying solely on Nvidia hardware. For Apple, it’s also a bet that the skills behind its consumer chips — the same silicon in the iPhone and Mac — can carry over into a business it walked away from a decade and a half ago.
Nothing here is close to shipping. But if Apple follows through, 2029 would mark the return of an Apple-branded box built to sit in someone else’s server rack.
Frequently Asked Questions
What is Apple’s M8 Ultra chip?
The M8 Ultra is an unreleased, unannounced Apple chip reportedly planned for a future enterprise AI server. Apple’s current top chip is the M5 Ultra; the M8 Ultra would arrive several generations later.
When will Apple’s AI servers launch?
According to The Information’s report, Apple isn’t expecting to launch the server before 2029, and the project’s plans aren’t finalized yet.
Will Apple’s AI servers use Nvidia technology?
Apple is reportedly considering Nvidia’s NVLink Fusion to connect multiple M8 Ultra chips inside each server. No partnership has been confirmed by either company.
Is this Apple’s return to the server business?
If it ships, yes. Apple last sold dedicated servers under the Xserve name, discontinued in January 2011.
