Breaking02 / IT
NVIDIA publishes guide for Python data reads into GPU memory
NVIDIA points to the efficiency of moving data from storage as compute and datasets grow together.

NVIDIA published a post on its developer blog on 23 Oct 2026 about using GPUDirect Storage, a technology that provides a direct data path between storage and GPU memory. The post applies to developers who read data into GPU memory from Python.
NVIDIA wrote at the start of the post that the datasets handled by GPU applications keep growing. Accelerators are becoming more powerful at the same time. As compute and data scale together, the post turns to how efficiently data moves from storage.
The company cited the 2025 Stanford AI Index for the pace of that growth. Compute for notable AI models has been doubling about every five months. The size of datasets used to train large language models (LLMs) has been doubling about every eight months.
The introduction to the post does not state which versions apply or which hardware is supported.