Apple Inc. (AAPL  ) is developing an AI server that could use two or four planned M8 Ultra chips, The Information reported on Wednesday. The iPhone company is considering Nvidia Corp's (NVDA  ) NVLink Fusion technology to connect the processors. The partnership could potentially integrate Apple's custom chips with NVIDIA's broader data-center systems.

In-House Silicon Push Expands

The project would extend Apple's chip strategy beyond consumer devices. Apple recently refreshed its Mac lineup with M6, M5 Pro, M5 Max and M5 Ultra chips, saying the 2-nanometer M6 delivers up to four times faster AI performance.

The server initiative reportedly began about a year ago with backing from current CEO John Ternus.

Apple does not expect a launch before 2029, and the project could still be canceled or proceed without NVIDIA technology.

If launched, it would mark Apple's first dedicated server hardware since it discontinued Xserve in 2011.

Broadcom Inc. (AVGO  ) could emerge as a key beneficiary as AI spending shifts from training models toward running large-scale inference workloads.

Lip Favors Broadcom For Inference

BakerAvenue Wealth Management chief strategist King Lip said Nvidia has excelled in AI model training. But for large, repetitive inference workloads, Broadcom is Lip's "top pick" for the next evolution in custom AI chips and inference play.

He added that Broadcom doesn't need Nvidia to lose market share to expand, as both companies can benefit from different parts of growing AI infrastructure demand.

Price Action: NVIDIA shares were up 1.89% at $216.17 and Apple shares were up 0.40% at $332.66 at the time of publication on Wednesday, according to Benzinga Pro data.