So I asked Bitget GetAgent for a full breakdown, and the answer was clearer than I expected.

The AI trade is no longer just about GPUs.

We’re entering the “memory bottleneck” phase.

Training and inference for LLMs now require massive amounts of:

HBM memory

NAND flash

SSD storage

Nearline HDDs

Data center bandwidth

And demand is growing faster than supply.

That’s why stocks like Micron ($MU), Sandisk ($SNDK), and Western Digital ($WDC) started going vertical.

A few things stood out to me:

• Micron’s HBM supply for 2026 is reportedly already sold out

• AI data centers are expected to consume most high-end DRAM production this year

• NAND and enterprise SSD demand keeps accelerating because inference workloads need huge storage capacity

• HDDs are unexpectedly back because AI infrastructure still needs cheap large-scale storage

This also explains why I’m watching:

$MU

$SNDK

$WDC

$ARM

$AMD

$QCOM

Most people focus only on compute.

But AI infrastructure has 3 layers:

Compute

Memory

Storage

And memory/storage might be the most underpriced part of the cycle right now.

My current positioning:

Core exposure: AI infrastructure

Aggressive exposure: memory + storage

Longer-term bet: edge AI via ARM/QCOM

Feels like the market is slowly realizing that without memory, AI scaling hits a wall.

[留言]

为什么值得关注

能改变理解方式，而不只是重复常识；符合当前抓取需求；它提供了新的理解或解释，而不只是表面观点

来源：reddit，领域：news，保留分：0.67