Open specs could cut AI data center hardware integration from a year to months

As AI factories scale, hardware interoperability is becoming as important as raw performance. Accelerators, memory and storage now come from many different suppliers, and every mismatch between them slows deployment across the data center. Pressure is mounting as AI infrastructure bottlenecks shift attention from individual chips to entire systems.