Sun Ge's saying "never enough storage" you need to listen to.

The storage sector may completely shed that "cyclical" skin and become a pure growth stock.
The core variable lies in AI agents.
In the past, when we used computers, closing the webpage freed up memory.
Now agents are "stateful" — they have to remember your habits, workflows, and historical conversations, all of which are firmly anchored in DRAM.
In the past, a server core with 4 GB of memory was the ceiling; from now on, the starting point will be 16 GB or even 32 GB.
Do the math: global DRAM production this year is only about 47 EB, but given the current expansion rate of CPU servers, by 2030 new demand alone could swallow 96 EB.
This doesn't even account for the impact of CXMT's expansion. Although many worry that CXMT and YMTC will disrupt the market, CXMT's density is only half that of the Big Three, so its impact on global capacity is only about 3% — barely a dent.
Many people worry that storage will fall, but there is actually a huge "reservoir" on the demand side.
As soon as memory prices drop, edge AI and AI PCs will immediately pull their configurations from 8 GB to 128 GB. This latent demand can firmly hold the price floor.
If you also think that computing power is just an appetizer and storage is the main course, forward this to friends who are still fixated on GPUs.
🤡
NVIDIA's current role is more like a "porter" for storage.
In the short term, it lacks chips; in the long term, it's working for storage.
Once the storage industry enters a super cycle, this boom will last at least five years.