TGViewer
Threading on the Edge Threading on the Edge @threadingontheedge · 7.78K subscribers
Post #4429 369
here my thesis on memory/storage as a subset of the ai supercycle

storage has had supercycles before. 2010. 2014. 2017.

the core trading logic is simple: AI devour compute power + voraciously consuming storage, with even 2026 capacity nearly fully pre-booked.

the 2026 $SNDK $MU $WDC $STX storage trade still has significant upside because:

- the storage supercycle lasts 3-5 years on current demand trajectory
GF Securities has a $1,462 target that was set conservatively before this week's volume data
- $SNDK $2,000 by year end implies ~40% upside on an asset with locked-in demand
- the late-cycle rotation from compute into storage has barely started institutionally

in crypto we will also see the same play, the crypto storage play $FIL $AR $STORJ is a different bet entirely. thesis is simple the move in equities will be reflected in that of the crypto markets

every storage supercycle was demand-driven by consumer devices: smartphones, PCs, streaming

the cycle would run beucase of AI, supply would catch up within 12-18 months, margins would compress, the trade would end

this cycle is different in 3 ways:

1. the demand source has a different cost tolerance
when a consumer buys a smartphone with 256GB, they're price-sensitive. they'll wait a cycle if prices spike.
when AWS pre-books 2026 storage capacity for AI infrastructure, they don't care if NAND (a type of memory chip used for storage) prices go up 40%. the margin on selling compute to enterprises covers it easily. they cannot afford to not have the storage.
demand is price-inelastic in a way consumer storage never was.

2. the product mix is shifting in favor of manufacturers
AI workloads specifically require high-density, high-endurance flash with specific write patterns. it's enterprise SSD with meaningfully higher average selling price and margins.
SanDisk, Micron, and Western Digital all have better product mix today than any prior cycle. they're selling $300 enterprise SSDs.

3. supply cannot catch up in 12 months anymore
advanced NAND fab construction timelines are now 3-5 years, not 18 months. equipment lead times, cleanroom build time, and process qualification have all extended.
if you pre-book 2026 capacity now, you're acknowledging that you cannot build or buy your way to more supply on a short timeline.
this cycle does not end the way prior cycles ended.

also, oil at $200 also accelerates the energy efficiency mandate for data centers which directly benefits high-density, low-power enterprise NAND over spinning disk

capital is rotating out of @nvidia. what’s interesting is that $NVDA has been down for several days, while others like
$SNDK $MU $WDC $STX are still pushing higher.


https://x.com/arndxt_xo/status/2051907255373910066
X (formerly Twitter) arndxt (@arndxt_xo) on X here my thesis on memory/storage as a subset of the ai supercycle storage has had supercycles before. 2010. 2014. 2017. the core trading logic is simple: AI devour compute power + voraciously consuming storage, with even 2026 capacity nearly fully pre-booked.…
  • 👍 2
More from @threadingontheedge
  1. Oct 2, 2026Aave v3 Loop Safe Module Loss: ~114.09 ETH https://x.com/SlowMist_Team/status/210585527653…
  2. Oct 1, 2026>sell the news over >next leg soon >rh szn https://x.com/tombxbt/status/210561976517602526…
  3. Oct 1, 2026Crypto + Macro Stuff I'm Looking At Today https://x.com/rektdiomedes/status/21053987511333…
  4. Oct 1, 2026The birthplace of DeFi is shutting down! Synthetix Discord was THE PLACE to be in 2019. So…
  5. Oct 1, 2026Robotics sector could print many new millionaires in the next 5 years.💰 https://x.com/Cry…
  6. Oct 1, 2026If you're going in as a contracts trader, you're facing a casino like this: Zero front-end…
Threads Profile ViewerView any public Threads profile without an account.Open ThreadLook →Writing with AI? Make it sound human.Metric37 rewrites AI drafts so they read naturally. Free AI detector, 1,500 words free.Try Metric37 →