AI and Storage: The Hidden Challenge Behind the Race for Performance

The Hidden Face of the AI Revolution: Data Storage

In the world of artificial intelligence, discussions often revolve around computing power, high-performance GPUs, and gigantic data centers. However, a crucial yet less visible aspect of this technological revolution is data storage. Indeed, AI does not just consume data; it produces and reuses it exponentially.

A Growing Storage Challenge

With the rise of generative AI, investments have focused on developing powerful processors capable of handling complex models. However, a critical question arises: where to store the enormous volumes of data necessary for the operation of these models?

Nicolas Frapard, Senior Sales Director for Europe at Western Digital, emphasizes that storage is often undervalued in the industry. According to him, "AI is fundamentally a data system. Computing processes this data, but it persists and accumulates, making storage indispensable and requiring sustainable and economically viable evolution."

As companies move from the experimentation phase to large-scale deployment, the importance of storage becomes undeniable. Each new model version and every query generates data that must be preserved, managed, and accessible.

The Continuous Accumulation of Data

Generative AI has transformed storage needs, forcing companies to manage not only their historical data but also the continuously created data by the models. This includes logs, metadata, as well as results and synthetic training data.

Nicolas Frapard explains that "generative AI redefines storage needs in terms of scale and complexity of the lifecycle. The challenge is how to effectively manage, prioritize, and reuse data over time."

Despite this, companies often continue to prioritize investments in computing at the expense of their data infrastructure. "Investment decisions are still largely driven by performance, as they provide immediate and visible results. Storage often comes later, even though it is the foundation of the entire system."

The danger lies in building high-performance architectures in the short term that are unable to sustainably manage data growth. At scale, resilience, lifecycle management, and economic efficiency become as critical as performance itself.

The Importance of Hard Drives

In this context, hard drives play a central role in the AI economy. According to IDC, about 80% of data in cloud environments is still stored on hard drives (HDD). This is primarily due to cost.

"HDDs remain essential for storing large volumes of data economically and at scale. AI is primarily a data challenge, not just a matter of performance or cost. While the cost per terabyte is an advantage, the true value of hard drives lies in their ability to optimize the total cost of ownership at scale."

Western Digital emphasizes that Flash storage is crucial for high-performance applications, but it remains more expensive for mass storage. New generations of hard drives also improve the energy efficiency of infrastructures. Moving from 26 TB to 32 TB to deploy an exabyte of storage can reduce the number of drives needed by 18.8%, leading to lower electricity consumption and cooling needs.

A Strategic Issue for Europe

For Nicolas Frapard, the issue of storage goes beyond the technical framework and touches on digital sovereignty. "Europe has strong capabilities in AI innovation, but there is a gap in large-scale infrastructures. A significant portion of storage capacity is concentrated outside the region."

According to Western Digital, mastering data infrastructures is crucial for enabling European companies to retain, govern, and utilize their data locally, in compliance with regulatory requirements and digital sovereignty ambitions. As data volumes explode, the ability to store, organize, and leverage this data could become a competitive advantage as decisive as computing power itself.