Jukan: AI model inference performance relies more on memory than Nvidia GPU

Citrini Research analyst Jukan stated that improving inference performance is not solely achieved by adding Nvidia GPUs. GPUs are often idle due to memory bottlenecks in inference, and the return on investment for inference depends on memory rather than GPUs. The importance of Micron may surpass Nvidia.