Leaked photos have offered a rare glimpse inside what appears to be Apple's Private Cloud Compute server hardware, suggesting it runs on the company's own M5 silicon.
Images shared by X user @hsuchingpo show a rack-mount enclosure packed with multiple compute modules.
Each module sits on a PCIe-style card with blue ejector latches, clearly designed for hot-swapping by technicians.
A service diagram visible in the photos lists 14 numbered board positions, each with its own Apple part number (APN).
A close-up of one module shows a single M5 package under thermal paste, with VRM circuitry along the lower edge of the board and a gold edge connector that plugs into the backplane.
The board itself has an Apple-style part-number sticker that reads 631-13397.
A wider shot of the chassis reveals a fully populated server blade, with compute modules lined up across carrier bars.
Each bar holds eight boards, and two rows are visible, connected by a central heat-pipe manifold.
The service diagram taped inside the lid lists positions 0 through 13 with unique APNs, pointing to a modular setup meant for field replacement rather than a fixed, single-board design.
It's possible the full chassis can hold more modules than the two rows shown, or the diagram might not align perfectly with a symmetrical layout.
While the M5 isn't as powerful as Apple's higher-end workstation chips like the M3 Ultra in raw multi-core processing, its higher efficiency allows multiple units to be packed densely into one enclosure without major thermal issues.
Apple's Private Cloud Compute setup relies on device-class silicon instead of traditional server processors, essentially extending the company's on-device privacy approach to larger-scale AI work in the cloud.
Although the photos could show an engineering or pre-production configuration, they provide one of the clearest public looks yet at how Apple may be building its secure cloud AI systems.
