苹果最强芯片 M8 Ultra，能让 Mac 重回服务器市场吗？

Apple is developing an AI server featuring two or four M8 Ultra chips interconnected via Nvidia's NVLink, targeting external customers with a potential 2029 launch, though the project remains unconfirmed. The move follows surging enterprise demand for Mac mini/Mac Studio deployments by labs like OpenAI and Anthropic - driven by unified memory advantages (up to 512GB at 1.2TB/s on M5 Ultra) and macOS licensing constraints - which boosted Mac revenue 29% quarterly. However, Apple's 2011 Xserve failure and its consumer-focused business model complicate the effort; the server may instead ship as a high-density AI inference appliance leveraging mature private-cloud-computing (PCC) technology, with rack design and maintenance outsourced to third-party hosts.

why now

Mac mini需求暴涨涨价，AI推理正热，企业市场机不可失。

topic

AI Industry Impact

source

爱范儿