When AI Becomes an 'Agent,' Which Chips Win? — The Rise of eSSD and NAND

AI Chips Aren't Just About HBM Anymore

When you think of AI chips, SK Hynix's HBM (High Bandwidth Memory — the high-performance memory used for AI computation) probably comes to mind first. But the next chip category getting serious attention in the AI world is NAND flash — the storage semiconductor found in smartphones and SSDs. So why is a storage chip, not a computing chip, suddenly stealing the spotlight?

AI Is Shifting from 'Studying' to 'Working'

Until now, AI has been mostly in the 'training' phase — feeding in massive amounts of data and learning patterns. Raw computation speed was everything in that phase, which is why HBM was king. But AI is now moving into the 'Agent' phase. An agentic AI doesn't just learn — it makes decisions and takes action on its own. It searches the web, writes emails, and generates reports. Every one of those tasks requires loading and saving enormous amounts of data in real time.

What Is an eSSD?

An eSSD is an Enterprise SSD — the same basic technology as the SSD in your laptop, but built much larger and faster for servers and data centers. A single AI agent can process more data in a day than most people handle in a lifetime. Storing and retrieving that data at speed requires high-performance eSSD drives deployed at massive scale.

What Does This Mean for Samsung and SK Hynix?

Both Samsung Electronics and SK Hynix manufacture NAND flash. If the agentic AI era takes hold, the NAND business at both companies could grow just as powerfully as their HBM business. Industry forecasts suggest NAND makers could post all-time record revenues in the second half of this year. If you've been tracking AI chip exposure only through HBM, NAND is now a category worth understanding alongside it.