Apple Reportedly Building an Enterprise AI Server With Its Own M8 Ultra Chips, Eyeing 2029

Apple hasn't sold a server since the Xserve got discontinued in January 2011. That's about to change, if a report from The Information holds up.

According to The Information, Apple is developing an enterprise AI server built around its own future M8 Ultra chips. The plan reportedly calls for two configurations: a smaller unit with two M8 Ultra chips clustered together, and a bigger one running four. Apple would sell these to AI developers, businesses, and governments, not just use them in-house.

Don't expect one in a data center anytime soon. Multiple outlets covering The Information's reporting, including Ars Technica and The Decoder, note a target launch of 2029 at the earliest, and the whole thing could still get scrapped before then.

Ternus's Fingerprints Are On This

The project reportedly got its green light about a year ago from John Ternus, who was then running Apple's hardware engineering division. Ternus officially became Apple's CEO on September 1, 2026, taking over from Tim Cook after Cook's 15-year run atop the company, according to Ars Technica. If this server ships, it's one of the first big bets tied to his tenure.

Apple is also reportedly in talks with Nvidia about using its NVLink Fusion technology to connect the M8 chips inside the server. NVLink Fusion is Nvidia's interconnect system, already licensed out to Qualcomm, Arm, Fujitsu and Marvell, per reporting from The Next Web. The News International notes Nvidia pulls in roughly a fifth of its data-center revenue from networking hardware like this, meaning an Apple deal would hand Nvidia a marquee customer even without selling a single GPU. The Next Web reported that shares of both Apple and Nvidia barely moved on the news.

Why Now

The timing isn't random. AI labs are already buying Apple's existing hardware in bulk. Per The Information's reporting, OpenAI has purchased "tens of thousands" of Mac minis and Mac Studios to train AI agents through reinforcement learning, and Anthropic rents Mac minis through Amazon Web Services. That demand shows up in Apple's earnings: Mac revenue jumped nearly 29 percent to $10.4 billion in Apple's most recent quarter, according to The News International and The Decoder, making Mac the company's fastest-growing product line.

That surge is also colliding with a supply problem the whole tech industry is fighting. Ars Technica points to an ongoing memory chip shortage driven by the AI data-center buildout, which has already pushed up prices on gaming devices, flagship phones, and a range of Apple products. Whatever Apple builds, it'll need memory chips it may struggle to get cheaply.

The Europe Wrinkle

There's a government-procurement angle worth noting. The Next Web reported that the European Commission proposed a Cloud and AI Development Act in June 2026 that would grade public cloud contracts by how independent they are from non-European control, with the strictest tier reserved for defense and national security work requiring full European ownership of the software stack.

Sovereignty advocates in Europe have a legitimate worry: handing critical government AI infrastructure to an American company's proprietary hardware, especially paired with Nvidia's networking tech, means a foreign company controls the design, maintenance, and supply chain of systems that could touch national security.

But the Act as written covers cloud services, not hardware purchases, so an outright Apple server sale would fall outside it entirely, according to The Next Web. And Europe's own procurement record undercuts the purity argument: a six-year, €180 million sovereign cloud contract awarded in April 2026 went partly to a consortium built on American infrastructure, including a Google-linked venture, alongside France's Mistral AI. The Commission itself said at the time that "non-European technologies, operated within a strict and appropriate framework, can meet the minimum level of sovereignty required." Regulation hasn't stopped Europe from buying American silicon when it wants to; it's just added paperwork.

Apple has said nothing publicly about any of this. The Information's sourcing describes an internal project still years from market, one that could ship with Nvidia's tech, without it, or not at all. Whether Ternus keeps pushing it past prototype stage, and whether Apple can even get enough memory chips to build it at scale by 2029, remains an open question nobody at Apple has answered on the record.