Apple is reportedly planning a return to the server business, and it may bring an unlikely partner along for the ride: Nvidia. According to The Verge AI, citing a report from The Information, the company is weighing a new enterprise machine built around its ARM-based M-series silicon, with the first product not expected until 2029.
Threat assessment for the incumbents: low today, real by decade’s end. Opportunity assessment for Apple: the AI compute shortage is the biggest hardware opening the company has seen since the iPhone, and it currently has zero products aimed at it.
What we know
The details are thin, but the shape is clear. Here’s what The Verge AI reports:
- Apple retired its Xserve line in 2011 and has ignored the enterprise rack ever since.
- The new server, whatever it ends up being called, would run on two or four M8 Ultra chips depending on the model.
- Apple only announced the M6 in August, so M8 is at least two generations away. That explains the 2029 target.
- The machine may use Nvidia’s NVLink Fusion, an interconnect that lets multiple chips behave as a single unit.
- Plans could still change. This is a report about intentions, not a product announcement.
Why Apple thinks it can win here
The logic runs through the Mac Mini and Mac Studio. Both machines have become surprisingly popular with AI developers, popular enough to cause shortages, as The Verge AI notes. The reason is unified memory. Apple’s M-series chips share one large pool of RAM between CPU and GPU, which means a Mac Studio can load a large language model that would choke a consumer Nvidia card. Developers figured this out on their own. Apple appears to have noticed.
What stands out here is the efficiency angle. Data centers are hitting power limits before they hit budget limits. If Apple can deliver strong inference performance per watt, it has a pitch that doesn’t require beating Nvidia on raw throughput. It just needs to be cheaper to run.
The Nvidia angle is the real story
Apple and Nvidia have barely spoken for over a decade. The relationship soured after GPU failures in MacBooks during the late 2000s, and Apple has shipped its own graphics ever since. So the idea of Apple servers using Nvidia interconnect tech is notable on its own.
But there’s precedent. The Verge AI points out that Apple recently turned to Nvidia chips to power its revamped Siri servers. That was a quiet admission that Apple’s own silicon couldn’t handle the load alone. NVLink Fusion in a future Apple server would extend that cooperation from “we buy your chips” to “we build around your standard.”
For Nvidia, this is a small but strategic win. NVLink Fusion is Nvidia’s play to stay at the center of the data center even when the main processor isn’t theirs. Getting Apple on board would validate that approach with the one chip company that has never needed Nvidia for anything.
What practitioners should take from this
Nothing changes for your infrastructure this year or next. But three things are worth tracking:
- If you’re running local inference on Mac Studios today, Apple now has a reason to keep improving that path rather than treating it as a happy accident.
- The interconnect standard war is heating up. NVLink Fusion versus UALink versus whatever Apple might have built alone matters for anyone planning multi-chip deployments after 2028.
- Enterprise Mac support, which Apple has treated as an afterthought, would have to get serious. Watch for hints in macOS server tooling over the next two years.
The honest caveat
A 2029 product is three chip generations away. Apple kills projects. Nvidia partnerships fall apart. The Verge AI is careful to frame this as reporting on plans, and so should you. The company has been rumored to be building AI servers before, and those rumors have mostly pointed at internal use for Apple Intelligence, not products for sale.
Still, the direction matters more than the date. Apple sat out the first AI hardware boom entirely. This report suggests it doesn’t intend to sit out the second one. More details are in The Verge AI’s original coverage.