The thesis is specific: memory chips — DRAM and especially high-bandwidth memory (HBM) — are entering an upcycle, and the companies that make them capture the pricing power. This is distinct from the broad AI or GPU trade. AI accelerators need vast amounts of memory stacked next to them, and memory is famously cyclical, so when demand outruns supply, the makers see prices and margins surge. This post walks through expressing that specific view on Legend. It is one applied example of the thesis-trade framework.
The Thesis
Memory is a commodity-like business that swings between glut and shortage. In a shortage, contract prices for DRAM and HBM climb, and the handful of makers that control supply enjoy expanding margins. The current driver is AI: every accelerator needs HBM, and high-end servers need ever more DRAM. If you believe the upcycle has room to run, you want long exposure to the memory makers specifically — not the GPU designers, who are a different layer of the AI trade. Legend gives you equity perps on the memory names in one self-custody account, and because stock perps trade around the clock you are not boxed into the regular session — here is why stock perps trade 24/7.
How to Express It on Legend
Go long the memory makers (basket)
The memory industry is concentrated in a few players. Build long positions across them:
- MU — Micron, the US DRAM and HBM maker
- SNDK — SanDisk, NAND flash storage
- KIOXIA — Japanese NAND flash maker
- SKHX — SK Hynix, a leading HBM supplier
- SMSN — Samsung, the largest memory maker
Spreading across the basket means a single company's execution miss hurts less than concentrating in one ticker. This is the diversified expression of the memory view.
Go long a single high-conviction name
If you have one standout, go long it alone — for many traders that is MU or SKHX, the names most levered to HBM. Highest reward if you are right, highest concentration risk if that one company stumbles.
Go long the DRAM index
If you like the theme but do not want to pick the winner, go long DRAM, the memory index perp. One position spreads exposure across the sector — the basket expression in a single instrument.
Relative value: long memory, short a lagging chip name
If your view is that memory outperforms the rest of semis, you can run a relative-value spread: long a memory name and short a semiconductor name you expect to lag, sized notional-balanced so broad sector moves largely offset.
Start trading on Legend to put any of these on as real positions.
Sizing and Risk
- Respect the cyclicality. Memory is the most cyclical corner of semis; size for sharp reversals, not just the upcycle. See how to manage risk.
- Decide your max loss before you size. These names are volatile around pricing data and guidance.
- Use leverage deliberately. Available leverage is a ceiling, not a recommendation. Some equity perps are isolated-margin-only — use that to cap the loss per leg.
- Watch the basket correlation. Memory names move together on industry pricing, so a basket is less diversified than it looks.
What Could Go Wrong
The memory supercycle trade rides a famously brutal cycle, and the turn can be fast. The main risks:
- Cycle rollover. Memory upcycles end in gluts; when supply catches up, prices and the stocks fall hard, often before the data confirms it.
- Earnings gaps. These names jump on guidance; a leveraged position can be liquidated on a single print. See how to avoid liquidation.
- Correlated drawdowns. The whole basket sells off together when the AI capex narrative wobbles.
- Wrong leg in a spread. A long-memory / short-laggard pair loses if the name you shorted outperforms.
This article is educational and is not financial advice.