Apple Wants Back in the Server Room, and Nvidia Might Help It Get There

Key points

• Apple is planning a return to the server market, possibly as early as 2029, according to The Verge, citing The Information.
• The servers are expected to run on two or four M8 Ultra chips, depending on the model.
• Apple announced its M6 chip in August 2026, putting the M8 generation at least a couple of product cycles away.
• Apple may use Nvidia's NVLink Fusion technology, which links multiple chips together so they behave as a single, more powerful unit.
• Apple's Mac Mini and Mac Studio have already become popular with AI developers, causing stock shortages.

Apple killed the Xserve in 2011 and left enterprise servers to everyone else. That might be about to change.

The reason isn't complicated. AI software needs enormous computing power, and Apple's M-series chips, the processors it designs for Macs and iPads, do a lot of work while using relatively little electricity. That combination has made the Mac Mini and Mac Studio genuinely popular among developers building AI tools, popular enough to cause shortages. Apple has apparently decided demand is big enough to warrant a dedicated server product.

What would these servers actually be?

The new machine would run on Apple's M8 Ultra chip, which doesn't exist yet. Apple announced the M6 in August 2026, and chip generations tend to come in pairs, so the M8 is probably three or four years out. That lines up with the rumoured 2029 launch window.

Depending on the model, the server could hold two or four of those chips working together. More chips means more work handled at once.

There's also talk of Apple using Nvidia's NVLink Fusion technology. NVLink Fusion connects separate chips with a fast link so they share memory and operate as one larger unit, which dramatically boosts performance for AI tasks. Apple and Nvidia haven't historically been close, but that's been shifting: Apple already uses Nvidia chips to run the servers behind its revamped Siri. We reported on NVLink Fusion when it first surfaced on our radar on 26 August 2026, and the Apple angle makes it considerably more interesting now.

Should anyone care about this before 2029?

For most people, not yet. This is early planning, and the product could change completely or not arrive at all.

The broader picture does matter now, though. If Apple pulls this off, businesses and AI developers get another option for powerful, energy-efficient server hardware, potentially at a price that competes with Dell or HP running Nvidia's own server chips.

The honest read here: this is Apple following the money. AI compute is where the business is going, Apple's chips are genuinely competitive on efficiency, and the Nvidia partnership is the thing worth watching most closely. Plans could still change, but the direction is clear.

Common questions

What was the Xserve?

The Xserve was Apple's rack-mounted server computer, sold mainly to businesses and universities. Apple discontinued it in January 2011 and hasn't sold dedicated server hardware since.

What does NVLink Fusion actually do?

NVLink Fusion, made by Nvidia, connects multiple processor chips with a fast link so they share memory and work together as one larger chip. It lets a server handle much bigger AI tasks than any single chip could manage alone.