Apple Reportedly Building M8 Ultra AI Server, in Talks with NVIDIA over NVLink Fusion

Per The Information, echoed by Mark Gurman, Apple is developing an enterprise AI server based on its own M-series silicon, with 2 or 4 M8 Ultra chip configurations, tentatively slated for 2029, and has approached NVIDIA about adopting NVLink Fusion interconnect. If true, it would mark Apple's return to the server market since Xserve was discontinued in 2011. Nothing is officially confirmed.

Confirmed

• The Information is the original source; @gappyvalley, @zephyrz9, @BenBajarin and others relayed highly consistent details
• @awnihannun notes Mark Gurman also reported the story, adding that Apple–NVIDIA relations have visibly warmed
• The server uses in-house M-series chips, in 2 or 4 M8 Ultra configurations, with NVIDIA NVLink Fusion interconnect under discussion
• Target customers are AI developers, enterprises and governments, offering private AI inference
• Mainstream reporting points to 2029 (@Polymarket's relay mentions "as early as 2026," conflicting with 2029)
• Apple's last server product was Xserve, discontinued in 2011
• A commenter on @awnihannun's post noted Apple already offers open-source tooling to run AI models efficiently on its hardware

Unconfirmed

• The NVIDIA cooperation is only at the discussion stage; final adoption of NVLink Fusion is undecided
• @asymco (Horace Dediu) relayed the rumor with the self-deprecating tag #ImSure, signaling skepticism; timelines cited range from 2026 to 2029
• Apple has made no official response

Why it matters

• If true, Apple would extend its custom silicon from consumer devices into the data center, formally entering the AI compute infrastructure race
• Pairing with NVIDIA networking suggests Apple may align with mainstream data-center interconnect ecosystems rather than going fully proprietary
• Private AI inference for enterprises and governments fits Apple's privacy positioning and could open a new enterprise revenue stream

2026-09-16 ~ 2026-09-18 · 8 related posts

• Apple reportedly building enterprise AI server on its own chips, launching 2029 — Polymarket · 2026-09-16
• Apple weighs return to server market, in talks with Nvidia for NVLink Fusion networking — zephyr_z9 · 2026-09-16
• [source] Apple Weighs M8-Based AI Server With Nvidia NVLink to Reenter Server Market — gappyvalley · 2026-09-16
• Apple reportedly exploring AI server return with M-series chips and NVDA NVLink Fusion — inductionheads · 2026-09-16
• Apple reportedly building enterprise AI servers on its own chips with Nvidia networking — asymco · 2026-09-17
• [source] Apple plots return to enterprise servers for first time since 2011, in talks with Nvidia on networking — awnihannun · 2026-09-17
• Apple Weighs AI Enterprise Servers With M8 Ultra Chips, Targeting 2029 — Beth_Kindig · 2026-09-18

1 near-duplicate retellings: BenBajarin