“…The slowdown in the rate of CPU scaling (which is evident in the extension in cloud service providers depreciation schedules from ~3 years pre-COVID to ~6 years today) continues to impact how and where data center capex dollars are spent and, specifically, in favor of accelerated computing. Looking ahead, with Nvidia driving a sustained reduction in the cost per unit of compute and, in turn, enabling the current transition from retrieval-based to generation-based Al technology, we believe the ongoing build-out of Gen Al infrastructure could extend beyond CY2025, particularly given the competitive dynamics that exist between competitors across most industry verticals.“
发布于 北京
