How will AI-focused data centers be different from traditional data centers? How are AI requirements reshaping the data center engineering discipline?

Dan Rabinovitsj elaborated on this in this short slip. Dan is VP of Infrastructure at Meta, leading a hardware engineering organization. Meta has been among the industry pioneers in AI. Apart from developing innovative Llama models, Meta has also invested in Infrastructure and data center capabilities required to train and use these models.

Key trends: Due to the AI requirements, the thermal density in the racks has been increasing significantly. Such power densification has been creating a lot of emphasis on liquid cooling instead of air cooling. Beyond these power and cooling challenges, almost every discipline of data center engineering needs to be rethought, given the scale of future AI data centers. Cc: AI at Meta.

Faisal Wattoo

Dan Rabinovitsj Meta AI at Meta

Silicon Valley Tech Talks