What Fat Option Premium Across the Memory Names Is Actually Pricing

Walk the memory names right now and the option premium looks generous almost everywhere. Cash-secured puts across the group annualise well into the double digits, and a few clear 50%. The natural instinct is to read that as opportunity. It pays to look closer first.

Premium gets rich when the options market is pricing a wide, uncertain range of outcomes. Fat premium across a whole cluster is that uncertainty handed back to you, quoted in basis points. It tells you about the range ahead, and it is worth understanding before you sell it.

There is a second thing worth knowing before you sell any of these: they do not all make the same thing. This is a moment-in-time slice inside the broader AI Buildout, and it spans three different businesses. Micron (MU) and SK Hynix (SKHY) make DRAM and HBM, the memory the AI shortage is actually about. SanDisk (SNDK) makes NAND flash and SSDs. Western Digital (WDC) and Seagate (STX) make hard-disk drives, the high-capacity storage that holds the data AI generates. One table, three different trades.

Why memory is the pinch-point now

For a while the bottleneck in the AI buildout was compute. Right now it is memory. Data centres are projected to take more than 70% of high-end memory output in 2026, and HBM revenue is tracking from roughly $35B in 2025 toward $60B in 2026. Contract prices rose again into the third quarter, DRAM up 13–18% quarter on quarter and NAND up 10–15%, and the more bullish analysts see no real relief before 2028.

That is a genuine demand shock. Layer it over memory's long history of boom and bust and you get an unusually wide range of possible outcomes. A wide range is what makes options premium fat, whether the width comes from a supercycle that keeps running or a glut that arrives early. The premium does not tell you which. It tells you the market cannot yet rule either out.

The storage names ride a related but separate driver. Servers need somewhere cheap to keep the flood of data AI produces, and that demand for high-capacity nearline disks is real and durable. It is not the HBM shortage, though, and it moves on its own cycle.

The one table that makes the point

Name	What it makes	Verdict	Discount*	Premium (annualised)	IV30	Value screen
MU	DRAM · HBM · NAND	wheelable	~30%	~28.7% (lowest)	94%	clears (cheapest)
SNDK	NAND flash	marginal	~33%	~52.6% (highest)	132%	0.00 (priciest)
WDC	HDD	wheelable	~22%	~38.9%	106%	0.33
STX	HDD	marginal	~23%	~40.6%	105%	0.33
SKHY	DRAM · HBM	not-wheelable	—	—	—	—

* Discount is your cost basis below the current price if assigned (the breakeven, which is the strike minus the premium). Premium is the annualised return on the collateral for the ~0.10-delta cash-secured put. At the implied vols in this table, a 0.10-delta put's real chance of finishing in the money is well above 10% — delta is not your chance of assignment, and the gap is widest exactly on names that pay like these. Illustrative, not a forecast. WDC and STX use a pre-earnings expiry (see below). Fact pack, as of 2026-07-11 ET.

The inversion

Read the table across and the premium and the valuation rank in opposite directions.

The richest premium in the group is SNDK, around 52.6% annualised at 132% implied volatility, and it sits on the name the value screen scores 0.00 after one of the largest runs in the market this year. The lowest premium of the tradeable four is MU, around 28.7%, and it sits on the only name that screens cheap. The market is paying the most to sell puts on the name it is least sure about, and the least on the one that looks most reasonably priced.

That is the useful read. Fat premium is the options market quoting a wide range. On its own it says nothing about whether the underlying is a name you would want to own. Same mechanism, very different companies underneath.

Name by name

MU (DRAM, HBM, NAND). Verdict: wheelable. It makes the memory the AI shortage is about, and it is the only name here that screens both reasonably priced and reasonable quality. Its ~0.10-delta put sits around the $705 strike, about 30% below the current price. Maximum loss about $69,005 (strike $705, credit $14.95); breakeven about $690.05. The catch is liquidity: open interest around 33 and a ~37% quoted spread even here, so size down and mind the fill. It also pays the least of the four, which is the trade-off for the least drama.

SNDK (NAND flash). Verdict: marginal. SanDisk spun out of Western Digital in 2025 and has since run roughly 720% (about 5,200% from the spin-off), which is why the value screen scores it 0.00 at a trailing P/E in the mid-60s. It pays the most by a mile: 132% implied vol, a put annualising about 52.6%, and a breakeven near $1,277, roughly 33% below spot. That looks like a large cushion, but the premium is the market pricing a genuinely wide range around a name priced for perfection. Maximum loss about $127,710 (strike $1,315, credit $37.90). Fine if you would happily own it a third lower, a warning if you would not. Street targets span $2,100 to $2,900, which mostly shows how little agreement there is.

WDC (HDD). Verdict: wheelable. After the SanDisk spin-off, Western Digital is a pure hard-disk-drive maker, the high-capacity nearline storage that holds AI's data. It reports on 29 July, so this pick steps back to the 24 July expiry to stay clear of the print. That keeps the trade clean, but a shorter 13-day tenor buys a closer strike, so the discount is smaller (about 22%) and the annualised premium higher (about 38.9%). Maximum loss about $45,363 (strike $460, credit $6.38); breakeven about $453.63. Liquidity is the thin part, with the widest quoted spread in the group.

STX (HDD). Verdict: marginal. Seagate makes the same enterprise nearline storage and rides the same real, separate storage-side driver. It reports on 28 July, so this pick expires 24 July, clear of the print. Clean, but short-dated and thinly quoted, with open interest in single digits, so size small and mind the fill. About 40.6% annualised; maximum loss about $70,465 (strike $715, credit $10.35); breakeven about $704.65, roughly 23% below spot.

SKHY (DRAM, HBM). Verdict: not-wheelable. SK Hynix is one of the world's two DRAM and HBM leaders, a core supplier to the very shortage this piece is about. Its Nasdaq ADR (each one a tenth of a Seoul-listed share) began trading on 2026-07-10 in a roughly $28B dual listing, one of the largest ever. One day into US trading it has no listed options, so there is no cash-secured put to price at any strike. Sometimes the premium you would want simply does not exist yet. Watch for options to be listed before treating it as a wheel candidate.

Three scenarios, briefly

Supercycle extends (~40%). HBM and DRAM stay tight into 2027. The more defensible angle is a far-out-of-the-money put on the name that screens cheap and reasonable, a smaller premium against a firmer floor. Invalidation: two straight quarters of falling contract prices, or visible HBM order cancellations.

Glut fear returns (~30%). Consumer demand is already cooling, and added capacity into a softening market is how prior cycles have peaked and reversed. This is the scenario the fat implied vol is largely pricing. A 0.10-delta put set well below the current price survives an ordinary pullback but not a full cycle turn.

The group splits (~30%). HBM and data-centre DRAM stay tight while commodity NAND and HDD storage move more on inventory and sentiment. Premium stays rich across the group, but only part of it is backed by a fair price. The same implied vol carries a different meaning name to name.

The honest read

Of this group, MU is the one name whose rich premium is backed by a fair valuation, and it carries no earnings event inside its expiry. Even there the option is thinly quoted, so size is the constraint. WDC and STX come out clean only because we step to a pre-earnings expiry, at the cost of a shorter tenor and a smaller discount. SNDK's premium is the richest precisely because the market is pricing the widest range around it. SKHY cannot be wheeled at all yet.

None of that is a recommendation. It is the premium, the maximum loss, and the breakeven laid side by side, so the largest number on the screen never travels alone. The lesson generalises past memory: when a whole group pays fat premium, look past the premium to what it is pricing. For the broader version of this idea, see why a cheap price is not a margin of safety, and the market regime you are selling into.

Method: reproducible fact pack (broker options quotes, yfinance fundamentals, as of 2026-07-11 ET); wheel scores and verdicts computed by levelbox.ai's own scorer. The yfinance fundamentals for this group looked patchy in places, so the quantitative claims here are anchored to the screen's computed wheel outputs rather than raw financial-statement line items.

Educational, not advice.