Apple returns to the server market in 2029 with M8 Ultra-based AI server

Summary

Apple is reportedly planning to sell a server to AI developers, enterprises, and governments starting in 2029. This M8 Ultra-based AI server, available in two configurations (2 or 4 chips), may utilize Nvidia's NVLink Fusion technology for inter-chip connectivity. This would mark Apple's first server sale since the Xserve was discontinued in 2011.

Why it matters

This move would symbolize Apple's return to the server market after two decades and pose a direct challenge to Nvidia's near-monopoly in AI training and inference infrastructure. Targeting governments is strategically significant, especially as the EU's new Cloud and AI Development Act phases public procurement based on sovereignty guarantees. Additionally, it indicates Apple's intent to extend the unified memory architecture advantage of consumer Macs into the enterprise AI computing sector.

Highlights

• Two different models: 2× M8 Ultra and 4× M8 Ultra; Nvidia NVLink Fusion is being considered for inter-chip connectivity but has not yet been confirmed
• The product will not be released before 2029; the project could be cancelled or proceed without using Nvidia technology
• Apple already runs Private Cloud Compute on Apple Silicon and has released Mac Studios with up to 512 GB of unified memory for local AI workloads

Critical take

The entire story relies on a single The Information report about an unannounced product that could still be cancelled. The emphasis on 'governments' is also weaker than it appears. The EU's Cloud and AI Development Act explicitly regulates cloud services, not hardware procurement. Therefore, selling a server box would fall entirely outside the law's scope. The near-zero market reaction in Apple and Nvidia stocks also indicates that investors view this as speculative news rather than a credible competitive threat in the near term.