Apple might make servers again to cash in on the AI rush

Apple left the server business when it retired Xserve in 2011. The Verge’s Terrence O’Brien, citing The Information, reports the company is planning a comeback aimed at AI compute, with a possible Nvidia pairing.

That matters because Apple’s ARM-based M chips already draw AI developers to Mac mini and Mac Studio—enough that shortages have shown up. If Cupertino builds rack gear around future M8 Ultra silicon, it isn’t just another Mac SKU. It’s Apple deciding the AI factory floor is too big a market to leave entirely to others.

The timeline is distant: the reported server likely wouldn’t debut until 2029, on two or four M8 Ultra chips, while Apple only just announced M6 in August. Rumors aren’t products. But the direction of travel is clear.

We’re reading this as Apple testing whether it can sell efficiency at scale the same way it sells it on the desk.

The Verge notes Mac mini and Mac Studio popularity with AI developers has led to shortages. That’s the demand signal. Apple’s answer, per The Information via O’Brien, is a server line years out—M8 Ultra configs, and possibly Nvidia’s NVLink Fusion so multiple chips act as one. The same piece says Apple recently turned to Nvidia chips for revamped Siri servers after a historically frosty relationship. Collaboration would be a strategic thaw, not nostalgia for Xserve branding.

Useful AI needs local and cloud compute that doesn’t waste watts. Apple’s pitch has always been performance-per-watt. If that lands in the rack, developers who already prototype on Macs get a path that stays inside Apple’s silicon story. We’re for that when it expands real capacity.

The harm path is lock-in dressed as convenience: a closed stack that prices out smaller labs, and a partnership that deepens Nvidia’s gravity in every serious cluster. Also: a 2029 debut means three years of rumor while competitors ship. Plans can change, as the Verge itself flags.

I’m watching whether NVLink Fusion shows up in any Apple disclosure before the M8 cycle, and whether developer shortages ease—or just migrate from desk to data center waitlists.

Apple’s last proper server line was Xserve, discontinued in 2011. Enterprise iron has since been Dell, HPE, Supermicro, and hyperscaler custom boards. Nvidia’s NVLink Fusion is the company’s interconnect play for tying chips into larger logical accelerators.