“Next-generation AI server racks contain approximately 120 FPGAs per rack.”