Apple Plans Return to Server Market, in Talks to Adopt NVIDIA Interconnect Technology
nashnova research
Apple is developing an AI server powered by its own M8 Ultra chips and is in discussions with Nvidia over NVLink Fusion interconnect technology — a move that redefines the boundary between cooperation and competition for two of tech's biggest rivals.
What does this server actually look like?
According to The Information, Apple plans two versions: a smaller unit with two M8 Ultra chips and a larger one with four M8 Ultra chips.
The M Ultra line is Apple's highest-performance silicon, already used in products like the Mac Studio. This means → Apple wants to stack consumer-grade chip performance into enterprise-class servers — a fundamentally different approach from Nvidia's purpose-built data-center GPUs.
Target customers include AI developers, enterprises, and government agencies. The earliest possible launch is 2029, but the project could still be canceled.
Why use Nvidia's technology?
Apple is considering Nvidia's NVLink Fusion — a package of switches, chiplets, and software that lets chips from different vendors communicate at high speed inside a data center.
In plain terms = Apple's own chips are powerful, but making multiple chips "talk to each other" efficiently is where Nvidia leads the industry. Some Apple engineers working on the project agree.
Apple already runs servers for its Private Cloud Compute system, but the in-house interconnect is considered slower and more expensive, making large-scale deployment difficult. This reflects a gap in Apple's data-center capabilities — precisely where Nvidia excels.
Mac sales are surging — what does that have to do with servers?
Over the past year, Mac mini and Mac Studio saw an unexpected demand spike among AI developers. Several major AI labs have purchased tens of thousands of these machines.
In its latest quarter, Mac revenue grew nearly 29% year-over-year to $10.4 billion, making it Apple's fastest-growing product line.
This means → the market has already voted with real money: Apple silicon can handle AI workloads. But Macs remain consumer-positioned, lacking enterprise features like remote management — and a dedicated server product is designed to fill that gap.
Aren't Apple and Nvidia old rivals?
The two companies' partnership dates back to 2001, when Steve Jobs called it "a great different relationship." It later soured over commercial disputes — reportedly in part because Jobs believed Nvidia's products infringed on technology from Pixar, the studio he founded.
Relations have warmed notably: while overhauling Siri, Apple began running some new AI features on Nvidia chips inside Google's data centers, which led to broader discussions between the two companies.
In plain terms = from years of silence to sitting down and discussing interconnect partnerships — the sheer demand for AI compute speaks louder than historical grudges.
What does this partnership mean for the market?
If Apple formally enters the AI server market, it will compete head-to-head on a track Nvidia dominates. Yet it would simultaneously rely on Nvidia's interconnect technology — a textbook "co-opetition" dynamic.
Apple's new CEO, John Ternus, is an early backer of the project. He took over earlier this month after leading Apple's hardware engineering division. This means → the project has top-level executive sponsorship inside Apple.
The key variable: whether the NVLink Fusion partnership ultimately materializes will determine whether Apple enters the data-center market as a collaborator or a pure competitor.
市场有风险，内容仅供研究参考，不构成投资建议。