ASIC makes it sound like it's a single chip, but in reality serving trillion-param models on Cerebras requires a full cluster (as in multiple racks, MW of power

ASIC makes it sound like it's a single chip, but in reality serving trillion-param models on Cerebras requires a full cluster (as in multiple racks, MW of power). Some interesting twitter analysis here: https://x.com/bleysg/status/2073937651150029084 https://x.com/bleysg/status/2073937651150029084