Apple May Return to Servers With Its Own AI Chips. Nvidia Could Still Get Paid

Apple may reintroduce AI servers, potentially incorporating its own M8 Ultra processors, according to a report from The Information on September 16. The architecture under consideration may involve two or four M8 Ultra chips, with a projected 2029 launch focused on AI inference. Nvidia's NVLink Fusion technology could be employed to connect these processors, but the decision regarding Apple's involvement remains undecided.

The proposed server architecture signifies a significant shift in AI hardware economics, as NVLink Fusion extends Nvidia's high-speed GPU communication strategy to custom CPUs and accelerators produced by other companies. This development could create an additional revenue stream for Nvidia, as networking equipment may account for 10% to 15% of AI data-center hardware expenses.

Conversely, the possibility of Apple competing with Nvidia using its own processors while still purchasing Nvidia's networking and software components introduces an intriguing scenario. Apple's M-series chips are already noted for their performance and power efficiency, which could enable the company to target AI developers, governments, and businesses seeking local inference capabilities.

However, the 2029 launch date presents challenges, including the potential for competitors to surpass Apple's advancements during the interim period. Moreover, Apple's return to the server market, which it exited in 2011, raises questions about the feasibility and timing of such a venture. Hedge fund activity reflects investor interest in both Apple and Nvidia, with Apple's short interest near 1% of its float and 3.5 days of trading volume.

While the server project may not materialize as initially reported, the strategic shift in AI hardware economics is an important development to watch.