[ewalk153]

Given these efficient numbers, I wouldn’t be surprised if Apple were racking Minis for serving the new ML models for Siri.

[favorited]

They've publicly disclosed that they built custom Apple Silicon servers to power Private Cloud Compute.

"The root of trust for Private Cloud Compute is our compute node: custom-built server hardware that brings the power and security of Apple silicon to the data center, with the same hardware security technologies used in iPhone, including the Secure Enclave and Secure Boot."

<https://security.apple.com/blog/private-cloud-compute/>

[wmf]

Mac minis probably don't have enough RAM to hold Siri's models. It's more likely that Apple is using (modified) Mac Studios.

[jsheard]

Why use Macs at all? It's well within Apples abilities to make a completely custom server motherboard built around Apple Silicon rather than hacking something together out of Mac parts. That would allow for much better rack density, and they could add proper server amenities like a BMC for remote management.

[opdahl]

You should check these guys out. They actually do it (rack, not hosting ML models) and it seems to me to be pretty elegant: https://www.macstadium.com/blog/m4-mac-mini-review