Micron's senior vice president, Jeremy Werner, told The Circuit Podcast that memory has become a strategic bottleneck for data-center inference, warning that insufficient memory can sharply cut GPU utilization while faster, larger memory can theoretically...The article requires paid subscription.Subscribe Now

Micron's senior vice president, Jeremy Werner, told The Circuit Podcast that memory has become a strategic bottleneck for data-center inference, warning that insufficient memory can sharply cut GPU utilization while faster, larger memory can theoretically multiply the compute extracted from GPUs. The remarks underscore how storage and memory design could limit AI deployment.