What AI lacks most right now might not be "brains," but "memory."

Stayed up late tonight listening to Micron's earnings report, and there was a very interesting term in it—"memory."

Back in school, I noticed that the truly smart classmates all had incredibly good memories: they wouldn't step into the same pit twice; they knew where to start with problems they'd seen before.

Human civilization is actually the same. The progress of civilization isn't just because human brains suddenly got smarter. It's because we've gotten better and better at preserving, replicating, and accessing memory.

From language, to writing, to books, to libraries, to the internet.

Each leap forward is about turning individual memory into collective memory.

But memory alone isn't enough; it also needs connection. This is what we've been calling "optical modules."

No matter how much of a genius Einstein was, he didn't build the entire cosmos by himself.

Behind general relativity, there's Riemannian geometry, the tensor tools of Ricci and Levi-Civita, and also Grossmann helping him translate physical intuition into mathematical language.

A single genius brain also needs to connect with another brain to become part of a larger network of civilization.

It's the same with AI. The GPU is responsible for thinking, memory is responsible for remembering, and interconnection is responsible for pulling that memory out.

Why is capital still chasing "memory"?

Why, even after seeing an 86% gross margin, are they still placing orders for Micron's memory?

This might not be simply speculating on a memory cycle, but betting on the next phase of AI—the memory system.

Can you imagine what kind of intelligence would erupt if one day, the world was truly moved into memory?

$Micron Tech(MU.US)$Marvell Tech(MRVL.US)$Roundhill Memory ETF(DRAM.US)$XL2CSOPHYNIX(07709.HK)$Roundhill T-Rex 2X Long DRAM Dly TrgtETF(RAM.US)