20 Comments
User's avatar
Neural Foundry's avatar

Outstanding comprehensiv breakdown of the memory supercycle dynamics. The Intel Optane section is particularly painful to read - they had first-mover advantage on persistent memory right as the AI wave was building and chose to exit. The DDR4 shortage angle is fascinating and underappreciated by most investors. Everyone talks about HBM but the cascading effects on conventional DRAM as fabs retool is creating opportunities across the whole memory stack. Your point about Sandisk's 250% run is well-taken, though I wonder if the QLC adoption curve for nearline storage could accelerate faster than the market expects, especially if video generation really takes off next year.

Expand full comment
Vikram Sekar's avatar

Thanks. I don't think the demand for NAND flash is going away in the next year or two. There is really a pressing need for storage in today's workloads.

Expand full comment
Paul Adal's avatar

Great read Vic.. is there still more upside in MU SNDK WDC STK?

Expand full comment
Vikram Sekar's avatar

I think the upcoming demand is partially priced in currently... but since I dont see the demand for storage or HBM going away any time soon... so yeah, I feel there is more to it.

Expand full comment
Paul Adal's avatar

which one of the 4 you think have the most upside from today

Expand full comment
Vikram Sekar's avatar

Can't say I know to be honest. Making stocks suggestions would make me investment advisory :D - bit more than I can handle there.

Expand full comment
SPF's avatar

Hi Vikram. Thanks for the information.

"Sandisk (SNDK+9.96%) has already seen its stock price skyrocket this year (250% up YTD) and this is even before the demand for high capacity storage has even taken off. "

Most of that growth has been in the last 30 days. Do you believe this is *not* the market pricing this exact AI video generation tailwind? If so, what do you attribute that gain to? Just picking your brain.

Expand full comment
Vikram Sekar's avatar

No I do think it is priced in, but since I dont see the demand going away for HBM or NAND, there is still future potential here. how much I cant tell.

Expand full comment
John's avatar

can't be fully priced in cause Sora 2 was out just few days ago.

Expand full comment
Vikram Sekar's avatar

It’s not just Sora, long HDD lead times are also known- and SSDs are a viable alternative

Expand full comment
SPF's avatar

Rumors have been out for a while tho. Industry insiders / analysts might have started buying.

Expand full comment
John's avatar

yes but how many thought it's going to be this good. Veo was ok, Vibes was terrible. Sora 2 exceeded everyone's expectations. just saying Sora 2 (and better and better AI videos from Veo and watever TikTok is doing) is another reason to be bullish on NAND

Expand full comment
SPF's avatar

Hmmm good point. OpenAI insiders would have known, but not that many people.

Expand full comment
Gossling's avatar

great analysis, thanks! i'm wondering if there is a picks-and-shovels strategy here. Micron just increased their capex guidance, who will profit?

Expand full comment
Vikram Sekar's avatar

Great question! I'll need to look into the supply chain closer to answer that. Nothing pops into my mind immediately at the moment.

Expand full comment
Rahul's avatar

Think LRCX, AMAT as their capex supply chain

Expand full comment
Peter W.'s avatar

I fully agree, especially with your take on Optane. IMHO a great example for Intel's seemingly infinite ability to kill off some of their great technologies just before they would have taken off and made them serious money. Optane was (is) amazing. The combination of almost DRAM-like performance (throughput, latencies) with NAND-like non-volatility however at incredible endurance would (I believe) still sell like hotcakes (or HBM3E 😀). I was especially impressed with the DRAM stick formfactor of Optane. Even before AI, it was very attractive for applications like HANA. And Optane drives are still sought-after as solid-state caches for HDD backups. Just try to buy some used ones!

Expand full comment
Vikram Sekar's avatar

Intels decisions have been truly stupid in the last decade or so

Expand full comment
michael's avatar

Bean counter mentality that see frontier tech as a cost and not an opportunity. I know a lot of people love Apple but they have had the same issue.

Expand full comment
Neural Foundry's avatar

Really insightful breakdown. On the storage side, Western Digital’s split and the Sandisk run you mentioned highlight how NAND dynamics can diverge from DRAM/HBM cycles. With QLC eSSDs creeping into nearline due to HDD supply tightness, WDC’s flash mix and controller IP could be a bigger lever than folks appreciate if AI video slop ramps in ’26. Curious if you see WDC’s nearline HDD roadmap (MAMR/ultraSMR) offsetting that SSD cannibalization, or will the elasticity tilt more decisvely to QLC?

Expand full comment